typefile
Object output with 3 properties
[
{
"avg_logprob": -0.24721175602504186,
"compression_ratio": 1.3163841807909604,
"end": 3.14,
"id": 0,
"no_speech_prob": 0.22674788534641266,
"seek": 0,
"start": 2.64,
"temperature": 0,
"text": " Okay.",
"tokens": [
50496,
1033,
13,
50521,
50640,
1144,
30569,
292,
13,
5198,
13,
8561,
13,
2555,
11,
16869,
11,
352,
2286,
13,
1044,
291,
11,
1074,
13,
50912,
50968,
865,
11,
3231,
13,
286,
478,
1391,
264,
1935,
13232,
954,
322,
341,
818,
11,
51272,
51356,
457,
286,
458,
264,
6583,
26107,
10137,
1074,
13,
286,
519,
321,
434,
411,
3356,
3521,
1314,
490,
1184,
661,
558,
586,
13,
51636,
51760
]
},
{
"avg_logprob": -0.24721175602504186,
"compression_ratio": 1.3163841807909604,
"end": 10.96,
"id": 1,
"no_speech_prob": 0.22674788534641266,
"seek": 0,
"start": 5.5200000000000005,
"temperature": 0,
"text": " Doxxed. Everyone. Cool. Please, Tyler, go ahead. Thank you, guys.",
"tokens": [
50496,
1033,
13,
50521,
50640,
1144,
30569,
292,
13,
5198,
13,
8561,
13,
2555,
11,
16869,
11,
352,
2286,
13,
1044,
291,
11,
1074,
13,
50912,
50968,
865,
11,
3231,
13,
286,
478,
1391,
264,
1935,
13232,
954,
322,
341,
818,
11,
51272,
51356,
457,
286,
458,
264,
6583,
26107,
10137,
1074,
13,
286,
519,
321,
434,
411,
3356,
3521,
1314,
490,
1184,
661,
558,
586,
13,
51636,
51760
]
},
{
"avg_logprob": -0.24721175602504186,
"compression_ratio": 1.3163841807909604,
"end": 18.16,
"id": 2,
"no_speech_prob": 0.22674788534641266,
"seek": 0,
"start": 12.08,
"temperature": 0,
"text": " Yeah, thanks. I'm probably the least intelligent person on this call,",
"tokens": [
50496,
1033,
13,
50521,
50640,
1144,
30569,
292,
13,
5198,
13,
8561,
13,
2555,
11,
16869,
11,
352,
2286,
13,
1044,
291,
11,
1074,
13,
50912,
50968,
865,
11,
3231,
13,
286,
478,
1391,
264,
1935,
13232,
954,
322,
341,
818,
11,
51272,
51356,
457,
286,
458,
264,
6583,
26107,
10137,
1074,
13,
286,
519,
321,
434,
411,
3356,
3521,
1314,
490,
1184,
661,
558,
586,
13,
51636,
51760
]
},
{
"avg_logprob": -0.24721175602504186,
"compression_ratio": 1.3163841807909604,
"end": 25.44,
"id": 3,
"no_speech_prob": 0.22674788534641266,
"seek": 0,
"start": 19.84,
"temperature": 0,
"text": " but I know the Modulus Lab guys. I think we're like 40 feet away from each other right now.",
"tokens": [
50496,
1033,
13,
50521,
50640,
1144,
30569,
292,
13,
5198,
13,
8561,
13,
2555,
11,
16869,
11,
352,
2286,
13,
1044,
291,
11,
1074,
13,
50912,
50968,
865,
11,
3231,
13,
286,
478,
1391,
264,
1935,
13232,
954,
322,
341,
818,
11,
51272,
51356,
457,
286,
458,
264,
6583,
26107,
10137,
1074,
13,
286,
519,
321,
434,
411,
3356,
3521,
1314,
490,
1184,
661,
558,
586,
13,
51636,
51760
]
},
{
"avg_logprob": -0.21478673856552333,
"compression_ratio": 1.4568527918781726,
"end": 33.84,
"id": 4,
"no_speech_prob": 0.00010068242409033701,
"seek": 2544,
"start": 25.44,
"temperature": 0,
"text": " So I built and run the program that they went through. They landed on Modulus Labs. But I guess",
"tokens": [
50364,
407,
286,
3094,
293,
1190,
264,
1461,
300,
436,
1437,
807,
13,
814,
15336,
322,
6583,
26107,
40047,
13,
583,
286,
2041,
50784,
50784,
472,
551,
286,
478,
1382,
281,
7019,
452,
1378,
926,
294,
341,
1901,
307,
11,
411,
11,
366,
456,
1629,
721,
300,
51292,
51292,
1176,
42,
12683,
393,
9528,
300,
500,
380,
4725,
2893,
264,
1101,
646,
12,
521,
1752,
420,
689,
11,
411,
11,
51700,
51700
]
},
{
"avg_logprob": -0.21478673856552333,
"compression_ratio": 1.4568527918781726,
"end": 44,
"id": 5,
"no_speech_prob": 0.00010068242409033701,
"seek": 2544,
"start": 33.84,
"temperature": 0,
"text": " one thing I'm trying to wrap my head around in this space is, like, are there certain things that",
"tokens": [
50364,
407,
286,
3094,
293,
1190,
264,
1461,
300,
436,
1437,
807,
13,
814,
15336,
322,
6583,
26107,
40047,
13,
583,
286,
2041,
50784,
50784,
472,
551,
286,
478,
1382,
281,
7019,
452,
1378,
926,
294,
341,
1901,
307,
11,
411,
11,
366,
456,
1629,
721,
300,
51292,
51292,
1176,
42,
12683,
393,
9528,
300,
500,
380,
4725,
2893,
264,
1101,
646,
12,
521,
1752,
420,
689,
11,
411,
11,
51700,
51700
]
},
{
"avg_logprob": -0.21478673856552333,
"compression_ratio": 1.4568527918781726,
"end": 52.16,
"id": 6,
"no_speech_prob": 0.00010068242409033701,
"seek": 2544,
"start": 44,
"temperature": 0,
"text": " ZKML can enable that don't necessarily provide the better back-end experience or where, like,",
"tokens": [
50364,
407,
286,
3094,
293,
1190,
264,
1461,
300,
436,
1437,
807,
13,
814,
15336,
322,
6583,
26107,
40047,
13,
583,
286,
2041,
50784,
50784,
472,
551,
286,
478,
1382,
281,
7019,
452,
1378,
926,
294,
341,
1901,
307,
11,
411,
11,
366,
456,
1629,
721,
300,
51292,
51292,
1176,
42,
12683,
393,
9528,
300,
500,
380,
4725,
2893,
264,
1101,
646,
12,
521,
1752,
420,
689,
11,
411,
11,
51700,
51700
]
},
{
"avg_logprob": -0.17045152640040917,
"compression_ratio": 1.6114285714285714,
"end": 59.44,
"id": 7,
"no_speech_prob": 0.00006204318924574181,
"seek": 5216,
"start": 52.16,
"temperature": 0,
"text": " oh, the tech makes this thing better, but where it provides, like, 10x better, like,",
"tokens": [
50364,
1954,
11,
264,
7553,
1669,
341,
551,
1101,
11,
457,
689,
309,
6417,
11,
411,
11,
1266,
87,
1101,
11,
411,
11,
50728,
50728,
917,
12,
18088,
420,
11,
411,
11,
9711,
1752,
30,
286,
2041,
341,
307,
445,
11,
411,
11,
257,
1154,
286,
536,
2108,
17240,
11,
51120,
51120,
689,
11,
411,
11,
286,
519,
439,
295,
505,
536,
264,
6228,
295,
17240,
11,
457,
321,
500,
380,
534,
536,
309,
885,
26336,
51404,
51404
]
},
{
"avg_logprob": -0.17045152640040917,
"compression_ratio": 1.6114285714285714,
"end": 67.28,
"id": 8,
"no_speech_prob": 0.00006204318924574181,
"seek": 5216,
"start": 59.44,
"temperature": 0,
"text": " end-user or, like, consumer experience? I guess this is just, like, a problem I see across crypto,",
"tokens": [
50364,
1954,
11,
264,
7553,
1669,
341,
551,
1101,
11,
457,
689,
309,
6417,
11,
411,
11,
1266,
87,
1101,
11,
411,
11,
50728,
50728,
917,
12,
18088,
420,
11,
411,
11,
9711,
1752,
30,
286,
2041,
341,
307,
445,
11,
411,
11,
257,
1154,
286,
536,
2108,
17240,
11,
51120,
51120,
689,
11,
411,
11,
286,
519,
439,
295,
505,
536,
264,
6228,
295,
17240,
11,
457,
321,
500,
380,
534,
536,
309,
885,
26336,
51404,
51404
]
},
{
"avg_logprob": -0.17045152640040917,
"compression_ratio": 1.6114285714285714,
"end": 72.96,
"id": 9,
"no_speech_prob": 0.00006204318924574181,
"seek": 5216,
"start": 67.28,
"temperature": 0,
"text": " where, like, I think all of us see the promise of crypto, but we don't really see it being readily",
"tokens": [
50364,
1954,
11,
264,
7553,
1669,
341,
551,
1101,
11,
457,
689,
309,
6417,
11,
411,
11,
1266,
87,
1101,
11,
411,
11,
50728,
50728,
917,
12,
18088,
420,
11,
411,
11,
9711,
1752,
30,
286,
2041,
341,
307,
445,
11,
411,
11,
257,
1154,
286,
536,
2108,
17240,
11,
51120,
51120,
689,
11,
411,
11,
286,
519,
439,
295,
505,
536,
264,
6228,
295,
17240,
11,
457,
321,
500,
380,
534,
536,
309,
885,
26336,
51404,
51404
]
},
{
"avg_logprob": -0.13395224677191842,
"compression_ratio": 1.5,
"end": 82.32,
"id": 10,
"no_speech_prob": 0.00030525089823640883,
"seek": 7296,
"start": 72.96,
"temperature": 0,
"text": " adopted. And so, yeah, one of the things I'm just trying to spend time on is where are the places",
"tokens": [
50364,
12175,
13,
400,
370,
11,
1338,
11,
472,
295,
264,
721,
286,
478,
445,
1382,
281,
3496,
565,
322,
307,
689,
366,
264,
3190,
50832,
50832,
689,
341,
393,
312,
6456,
300,
11,
291,
458,
11,
411,
11,
17879,
1319,
264,
636,
300,
1580,
576,
51340,
51340,
312,
1075,
281,
4648,
365,
341,
11311,
30,
286,
500,
380,
458,
498,
456,
311,
257,
665,
1867,
337,
300,
13,
51692,
51692
]
},
{
"avg_logprob": -0.13395224677191842,
"compression_ratio": 1.5,
"end": 92.47999999999999,
"id": 11,
"no_speech_prob": 0.00030525089823640883,
"seek": 7296,
"start": 82.32,
"temperature": 0,
"text": " where this can be applied that, you know, like, fundamentally change the way that someone would",
"tokens": [
50364,
12175,
13,
400,
370,
11,
1338,
11,
472,
295,
264,
721,
286,
478,
445,
1382,
281,
3496,
565,
322,
307,
689,
366,
264,
3190,
50832,
50832,
689,
341,
393,
312,
6456,
300,
11,
291,
458,
11,
411,
11,
17879,
1319,
264,
636,
300,
1580,
576,
51340,
51340,
312,
1075,
281,
4648,
365,
341,
11311,
30,
286,
500,
380,
458,
498,
456,
311,
257,
665,
1867,
337,
300,
13,
51692,
51692
]
},
{
"avg_logprob": -0.13395224677191842,
"compression_ratio": 1.5,
"end": 99.52,
"id": 12,
"no_speech_prob": 0.00030525089823640883,
"seek": 7296,
"start": 92.47999999999999,
"temperature": 0,
"text": " be able to interact with this ecosystem? I don't know if there's a good answer for that.",
"tokens": [
50364,
12175,
13,
400,
370,
11,
1338,
11,
472,
295,
264,
721,
286,
478,
445,
1382,
281,
3496,
565,
322,
307,
689,
366,
264,
3190,
50832,
50832,
689,
341,
393,
312,
6456,
300,
11,
291,
458,
11,
411,
11,
17879,
1319,
264,
636,
300,
1580,
576,
51340,
51340,
312,
1075,
281,
4648,
365,
341,
11311,
30,
286,
500,
380,
458,
498,
456,
311,
257,
665,
1867,
337,
300,
13,
51692,
51692
]
},
{
"avg_logprob": -0.19520848825437212,
"compression_ratio": 1.7148014440433212,
"end": 105.52,
"id": 13,
"no_speech_prob": 0.00009311757457908243,
"seek": 9952,
"start": 99.52,
"temperature": 0,
"text": " I have to jump in because this is actually why I started working on this was because I think it is",
"tokens": [
50364,
286,
362,
281,
3012,
294,
570,
341,
307,
767,
983,
286,
1409,
1364,
322,
341,
390,
570,
286,
519,
309,
307,
50664,
50664,
264,
733,
295,
264,
558,
3827,
13,
492,
434,
1237,
337,
257,
636,
281,
1065,
4562,
9769,
293,
281,
1969,
50900,
50900,
4069,
13952,
420,
1969,
4069,
2929,
1385,
11,
1936,
11,
1969,
527,
2035,
321,
362,
5191,
51120,
51120,
322,
5021,
1228,
733,
295,
544,
5164,
35799,
295,
15279,
300,
2878,
393,
1223,
11,
558,
30,
51372,
51372,
663,
366,
544,
411,
466,
577,
321,
434,
16741,
294,
4086,
11,
567,
321,
366,
11,
689,
321,
366,
11,
439,
264,
721,
51584,
51584
]
},
{
"avg_logprob": -0.19520848825437212,
"compression_ratio": 1.7148014440433212,
"end": 110.24,
"id": 14,
"no_speech_prob": 0.00009311757457908243,
"seek": 9952,
"start": 105.52,
"temperature": 0,
"text": " the kind of the right solution. We're looking for a way to own digital assets and to control",
"tokens": [
50364,
286,
362,
281,
3012,
294,
570,
341,
307,
767,
983,
286,
1409,
1364,
322,
341,
390,
570,
286,
519,
309,
307,
50664,
50664,
264,
733,
295,
264,
558,
3827,
13,
492,
434,
1237,
337,
257,
636,
281,
1065,
4562,
9769,
293,
281,
1969,
50900,
50900,
4069,
13952,
420,
1969,
4069,
2929,
1385,
11,
1936,
11,
1969,
527,
2035,
321,
362,
5191,
51120,
51120,
322,
5021,
1228,
733,
295,
544,
5164,
35799,
295,
15279,
300,
2878,
393,
1223,
11,
558,
30,
51372,
51372,
663,
366,
544,
411,
466,
577,
321,
434,
16741,
294,
4086,
11,
567,
321,
366,
11,
689,
321,
366,
11,
439,
264,
721,
51584,
51584
]
},
{
"avg_logprob": -0.19520848825437212,
"compression_ratio": 1.7148014440433212,
"end": 114.64,
"id": 15,
"no_speech_prob": 0.00009311757457908243,
"seek": 9952,
"start": 110.24,
"temperature": 0,
"text": " smart contracts or control smart wallets, basically, control our whatever we have activity",
"tokens": [
50364,
286,
362,
281,
3012,
294,
570,
341,
307,
767,
983,
286,
1409,
1364,
322,
341,
390,
570,
286,
519,
309,
307,
50664,
50664,
264,
733,
295,
264,
558,
3827,
13,
492,
434,
1237,
337,
257,
636,
281,
1065,
4562,
9769,
293,
281,
1969,
50900,
50900,
4069,
13952,
420,
1969,
4069,
2929,
1385,
11,
1936,
11,
1969,
527,
2035,
321,
362,
5191,
51120,
51120,
322,
5021,
1228,
733,
295,
544,
5164,
35799,
295,
15279,
300,
2878,
393,
1223,
11,
558,
30,
51372,
51372,
663,
366,
544,
411,
466,
577,
321,
434,
16741,
294,
4086,
11,
567,
321,
366,
11,
689,
321,
366,
11,
439,
264,
721,
51584,
51584
]
},
{
"avg_logprob": -0.19520848825437212,
"compression_ratio": 1.7148014440433212,
"end": 119.67999999999999,
"id": 16,
"no_speech_prob": 0.00009311757457908243,
"seek": 9952,
"start": 114.64,
"temperature": 0,
"text": " on chain using kind of more traditional notions of ownership that anyone can understand, right?",
"tokens": [
50364,
286,
362,
281,
3012,
294,
570,
341,
307,
767,
983,
286,
1409,
1364,
322,
341,
390,
570,
286,
519,
309,
307,
50664,
50664,
264,
733,
295,
264,
558,
3827,
13,
492,
434,
1237,
337,
257,
636,
281,
1065,
4562,
9769,
293,
281,
1969,
50900,
50900,
4069,
13952,
420,
1969,
4069,
2929,
1385,
11,
1936,
11,
1969,
527,
2035,
321,
362,
5191,
51120,
51120,
322,
5021,
1228,
733,
295,
544,
5164,
35799,
295,
15279,
300,
2878,
393,
1223,
11,
558,
30,
51372,
51372,
663,
366,
544,
411,
466,
577,
321,
434,
16741,
294,
4086,
11,
567,
321,
366,
11,
689,
321,
366,
11,
439,
264,
721,
51584,
51584
]
},
{
"avg_logprob": -0.19520848825437212,
"compression_ratio": 1.7148014440433212,
"end": 123.92,
"id": 17,
"no_speech_prob": 0.00009311757457908243,
"seek": 9952,
"start": 119.67999999999999,
"temperature": 0,
"text": " That are more like about how we're embedded in society, who we are, where we are, all the things",
"tokens": [
50364,
286,
362,
281,
3012,
294,
570,
341,
307,
767,
983,
286,
1409,
1364,
322,
341,
390,
570,
286,
519,
309,
307,
50664,
50664,
264,
733,
295,
264,
558,
3827,
13,
492,
434,
1237,
337,
257,
636,
281,
1065,
4562,
9769,
293,
281,
1969,
50900,
50900,
4069,
13952,
420,
1969,
4069,
2929,
1385,
11,
1936,
11,
1969,
527,
2035,
321,
362,
5191,
51120,
51120,
322,
5021,
1228,
733,
295,
544,
5164,
35799,
295,
15279,
300,
2878,
393,
1223,
11,
558,
30,
51372,
51372,
663,
366,
544,
411,
466,
577,
321,
434,
16741,
294,
4086,
11,
567,
321,
366,
11,
689,
321,
366,
11,
439,
264,
721,
51584,
51584
]
},
{
"avg_logprob": -0.18600391996079596,
"compression_ratio": 1.719298245614035,
"end": 131.12,
"id": 18,
"no_speech_prob": 0.0001739068393362686,
"seek": 12392,
"start": 123.92,
"temperature": 0,
"text": " that we could do, right? So after we've met, I often give the example after we've met and talked",
"tokens": [
50364,
300,
321,
727,
360,
11,
558,
30,
407,
934,
321,
600,
1131,
11,
286,
2049,
976,
264,
1365,
934,
321,
600,
1131,
293,
2825,
50724,
50724,
257,
857,
11,
286,
727,
980,
291,
11,
286,
727,
584,
11,
1767,
2845,
1266,
462,
9620,
281,
36983,
6254,
11,
558,
30,
400,
291,
1116,
312,
1075,
281,
360,
51088,
51088,
300,
6457,
731,
11,
291,
458,
11,
1310,
406,
322,
257,
960,
818,
365,
2452,
69,
3419,
293,
1507,
11,
457,
294,
954,
11,
291,
51356,
51356,
727,
360,
309,
11,
558,
30,
400,
370,
983,
393,
380,
321,
360,
300,
365,
11,
291,
458,
11,
983,
360,
321,
362,
281,
764,
2929,
1385,
293,
51584,
51584,
14093,
293,
1203,
30,
400,
300,
311,
445,
257,
17491,
6191,
1154,
13,
400,
498,
291,
519,
466,
341,
51840,
51840
]
},
{
"avg_logprob": -0.18600391996079596,
"compression_ratio": 1.719298245614035,
"end": 138.4,
"id": 19,
"no_speech_prob": 0.0001739068393362686,
"seek": 12392,
"start": 131.12,
"temperature": 0,
"text": " a bit, I could tell you, I could say, please send 10 ETH to Emiliano, right? And you'd be able to do",
"tokens": [
50364,
300,
321,
727,
360,
11,
558,
30,
407,
934,
321,
600,
1131,
11,
286,
2049,
976,
264,
1365,
934,
321,
600,
1131,
293,
2825,
50724,
50724,
257,
857,
11,
286,
727,
980,
291,
11,
286,
727,
584,
11,
1767,
2845,
1266,
462,
9620,
281,
36983,
6254,
11,
558,
30,
400,
291,
1116,
312,
1075,
281,
360,
51088,
51088,
300,
6457,
731,
11,
291,
458,
11,
1310,
406,
322,
257,
960,
818,
365,
2452,
69,
3419,
293,
1507,
11,
457,
294,
954,
11,
291,
51356,
51356,
727,
360,
309,
11,
558,
30,
400,
370,
983,
393,
380,
321,
360,
300,
365,
11,
291,
458,
11,
983,
360,
321,
362,
281,
764,
2929,
1385,
293,
51584,
51584,
14093,
293,
1203,
30,
400,
300,
311,
445,
257,
17491,
6191,
1154,
13,
400,
498,
291,
519,
466,
341,
51840,
51840
]
},
{
"avg_logprob": -0.18600391996079596,
"compression_ratio": 1.719298245614035,
"end": 143.76,
"id": 20,
"no_speech_prob": 0.0001739068393362686,
"seek": 12392,
"start": 138.4,
"temperature": 0,
"text": " that fairly well, you know, maybe not on a video call with deepfakes and stuff, but in person, you",
"tokens": [
50364,
300,
321,
727,
360,
11,
558,
30,
407,
934,
321,
600,
1131,
11,
286,
2049,
976,
264,
1365,
934,
321,
600,
1131,
293,
2825,
50724,
50724,
257,
857,
11,
286,
727,
980,
291,
11,
286,
727,
584,
11,
1767,
2845,
1266,
462,
9620,
281,
36983,
6254,
11,
558,
30,
400,
291,
1116,
312,
1075,
281,
360,
51088,
51088,
300,
6457,
731,
11,
291,
458,
11,
1310,
406,
322,
257,
960,
818,
365,
2452,
69,
3419,
293,
1507,
11,
457,
294,
954,
11,
291,
51356,
51356,
727,
360,
309,
11,
558,
30,
400,
370,
983,
393,
380,
321,
360,
300,
365,
11,
291,
458,
11,
983,
360,
321,
362,
281,
764,
2929,
1385,
293,
51584,
51584,
14093,
293,
1203,
30,
400,
300,
311,
445,
257,
17491,
6191,
1154,
13,
400,
498,
291,
519,
466,
341,
51840,
51840
]
},
{
"avg_logprob": -0.18600391996079596,
"compression_ratio": 1.719298245614035,
"end": 148.32,
"id": 21,
"no_speech_prob": 0.0001739068393362686,
"seek": 12392,
"start": 143.76,
"temperature": 0,
"text": " could do it, right? And so why can't we do that with, you know, why do we have to use wallets and",
"tokens": [
50364,
300,
321,
727,
360,
11,
558,
30,
407,
934,
321,
600,
1131,
11,
286,
2049,
976,
264,
1365,
934,
321,
600,
1131,
293,
2825,
50724,
50724,
257,
857,
11,
286,
727,
980,
291,
11,
286,
727,
584,
11,
1767,
2845,
1266,
462,
9620,
281,
36983,
6254,
11,
558,
30,
400,
291,
1116,
312,
1075,
281,
360,
51088,
51088,
300,
6457,
731,
11,
291,
458,
11,
1310,
406,
322,
257,
960,
818,
365,
2452,
69,
3419,
293,
1507,
11,
457,
294,
954,
11,
291,
51356,
51356,
727,
360,
309,
11,
558,
30,
400,
370,
983,
393,
380,
321,
360,
300,
365,
11,
291,
458,
11,
983,
360,
321,
362,
281,
764,
2929,
1385,
293,
51584,
51584,
14093,
293,
1203,
30,
400,
300,
311,
445,
257,
17491,
6191,
1154,
13,
400,
498,
291,
519,
466,
341,
51840,
51840
]
},
{
"avg_logprob": -0.18600391996079596,
"compression_ratio": 1.719298245614035,
"end": 153.44,
"id": 22,
"no_speech_prob": 0.0001739068393362686,
"seek": 12392,
"start": 148.32,
"temperature": 0,
"text": " secrets and everything? And that's just a purely technical problem. And if you think about this",
"tokens": [
50364,
300,
321,
727,
360,
11,
558,
30,
407,
934,
321,
600,
1131,
11,
286,
2049,
976,
264,
1365,
934,
321,
600,
1131,
293,
2825,
50724,
50724,
257,
857,
11,
286,
727,
980,
291,
11,
286,
727,
584,
11,
1767,
2845,
1266,
462,
9620,
281,
36983,
6254,
11,
558,
30,
400,
291,
1116,
312,
1075,
281,
360,
51088,
51088,
300,
6457,
731,
11,
291,
458,
11,
1310,
406,
322,
257,
960,
818,
365,
2452,
69,
3419,
293,
1507,
11,
457,
294,
954,
11,
291,
51356,
51356,
727,
360,
309,
11,
558,
30,
400,
370,
983,
393,
380,
321,
360,
300,
365,
11,
291,
458,
11,
983,
360,
321,
362,
281,
764,
2929,
1385,
293,
51584,
51584,
14093,
293,
1203,
30,
400,
300,
311,
445,
257,
17491,
6191,
1154,
13,
400,
498,
291,
519,
466,
341,
51840,
51840
]
},
{
"avg_logprob": -0.0912871031925596,
"compression_ratio": 1.7686832740213523,
"end": 158.4,
"id": 23,
"no_speech_prob": 0.0000475637789350003,
"seek": 15344,
"start": 153.44,
"temperature": 0,
"text": " model, like Celeste was talking about earlier, an ensemble of models, ZKML models that are measuring",
"tokens": [
50364,
2316,
11,
411,
19967,
8887,
390,
1417,
466,
3071,
11,
364,
19492,
295,
5245,
11,
1176,
42,
12683,
5245,
300,
366,
13389,
50612,
50612,
1203,
466,
291,
11,
1333,
295,
689,
291,
366,
11,
428,
5759,
11,
428,
11,
291,
458,
11,
428,
3256,
293,
3177,
293,
50880,
50880,
428,
4914,
11,
2503,
293,
439,
341,
733,
295,
1507,
11,
291,
727,
4728,
3811,
1419,
733,
295,
1266,
11,
1360,
51136,
51136,
5952,
26643,
11,
689,
291,
534,
362,
257,
588,
1090,
4314,
295,
6687,
300,
341,
307,
341,
51368,
51368,
954,
11,
293,
341,
307,
264,
8446,
436,
600,
12675,
13,
400,
550,
291,
14483,
300,
8446,
13,
400,
300,
727,
51612,
51612
]
},
{
"avg_logprob": -0.0912871031925596,
"compression_ratio": 1.7686832740213523,
"end": 163.76,
"id": 24,
"no_speech_prob": 0.0000475637789350003,
"seek": 15344,
"start": 158.4,
"temperature": 0,
"text": " everything about you, sort of where you are, your devices, your, you know, your image and voice and",
"tokens": [
50364,
2316,
11,
411,
19967,
8887,
390,
1417,
466,
3071,
11,
364,
19492,
295,
5245,
11,
1176,
42,
12683,
5245,
300,
366,
13389,
50612,
50612,
1203,
466,
291,
11,
1333,
295,
689,
291,
366,
11,
428,
5759,
11,
428,
11,
291,
458,
11,
428,
3256,
293,
3177,
293,
50880,
50880,
428,
4914,
11,
2503,
293,
439,
341,
733,
295,
1507,
11,
291,
727,
4728,
3811,
1419,
733,
295,
1266,
11,
1360,
51136,
51136,
5952,
26643,
11,
689,
291,
534,
362,
257,
588,
1090,
4314,
295,
6687,
300,
341,
307,
341,
51368,
51368,
954,
11,
293,
341,
307,
264,
8446,
436,
600,
12675,
13,
400,
550,
291,
14483,
300,
8446,
13,
400,
300,
727,
51612,
51612
]
},
{
"avg_logprob": -0.0912871031925596,
"compression_ratio": 1.7686832740213523,
"end": 168.88,
"id": 25,
"no_speech_prob": 0.0000475637789350003,
"seek": 15344,
"start": 163.76,
"temperature": 0,
"text": " your location, history and all this kind of stuff, you could eventually imagine having kind of 10,000",
"tokens": [
50364,
2316,
11,
411,
19967,
8887,
390,
1417,
466,
3071,
11,
364,
19492,
295,
5245,
11,
1176,
42,
12683,
5245,
300,
366,
13389,
50612,
50612,
1203,
466,
291,
11,
1333,
295,
689,
291,
366,
11,
428,
5759,
11,
428,
11,
291,
458,
11,
428,
3256,
293,
3177,
293,
50880,
50880,
428,
4914,
11,
2503,
293,
439,
341,
733,
295,
1507,
11,
291,
727,
4728,
3811,
1419,
733,
295,
1266,
11,
1360,
51136,
51136,
5952,
26643,
11,
689,
291,
534,
362,
257,
588,
1090,
4314,
295,
6687,
300,
341,
307,
341,
51368,
51368,
954,
11,
293,
341,
307,
264,
8446,
436,
600,
12675,
13,
400,
550,
291,
14483,
300,
8446,
13,
400,
300,
727,
51612,
51612
]
},
{
"avg_logprob": -0.0912871031925596,
"compression_ratio": 1.7686832740213523,
"end": 173.52,
"id": 26,
"no_speech_prob": 0.0000475637789350003,
"seek": 15344,
"start": 168.88,
"temperature": 0,
"text": " factor authentication, where you really have a very high degree of confidence that this is this",
"tokens": [
50364,
2316,
11,
411,
19967,
8887,
390,
1417,
466,
3071,
11,
364,
19492,
295,
5245,
11,
1176,
42,
12683,
5245,
300,
366,
13389,
50612,
50612,
1203,
466,
291,
11,
1333,
295,
689,
291,
366,
11,
428,
5759,
11,
428,
11,
291,
458,
11,
428,
3256,
293,
3177,
293,
50880,
50880,
428,
4914,
11,
2503,
293,
439,
341,
733,
295,
1507,
11,
291,
727,
4728,
3811,
1419,
733,
295,
1266,
11,
1360,
51136,
51136,
5952,
26643,
11,
689,
291,
534,
362,
257,
588,
1090,
4314,
295,
6687,
300,
341,
307,
341,
51368,
51368,
954,
11,
293,
341,
307,
264,
8446,
436,
600,
12675,
13,
400,
550,
291,
14483,
300,
8446,
13,
400,
300,
727,
51612,
51612
]
},
{
"avg_logprob": -0.0912871031925596,
"compression_ratio": 1.7686832740213523,
"end": 178.4,
"id": 27,
"no_speech_prob": 0.0000475637789350003,
"seek": 15344,
"start": 173.52,
"temperature": 0,
"text": " person, and this is the intent they've expressed. And then you execute that intent. And that could",
"tokens": [
50364,
2316,
11,
411,
19967,
8887,
390,
1417,
466,
3071,
11,
364,
19492,
295,
5245,
11,
1176,
42,
12683,
5245,
300,
366,
13389,
50612,
50612,
1203,
466,
291,
11,
1333,
295,
689,
291,
366,
11,
428,
5759,
11,
428,
11,
291,
458,
11,
428,
3256,
293,
3177,
293,
50880,
50880,
428,
4914,
11,
2503,
293,
439,
341,
733,
295,
1507,
11,
291,
727,
4728,
3811,
1419,
733,
295,
1266,
11,
1360,
51136,
51136,
5952,
26643,
11,
689,
291,
534,
362,
257,
588,
1090,
4314,
295,
6687,
300,
341,
307,
341,
51368,
51368,
954,
11,
293,
341,
307,
264,
8446,
436,
600,
12675,
13,
400,
550,
291,
14483,
300,
8446,
13,
400,
300,
727,
51612,
51612
]
},
{
"avg_logprob": -0.06850418677696815,
"compression_ratio": 1.7304964539007093,
"end": 186.32,
"id": 28,
"no_speech_prob": 0.00005061955380369909,
"seek": 17840,
"start": 178.4,
"temperature": 0,
"text": " all happen in sort of an automated way. So I think that that would be the huge win for the, for the",
"tokens": [
50364,
439,
1051,
294,
1333,
295,
364,
18473,
636,
13,
407,
286,
519,
300,
300,
576,
312,
264,
2603,
1942,
337,
264,
11,
337,
264,
50760,
50760,
9711,
293,
337,
12708,
295,
764,
13,
583,
309,
311,
1391,
588,
1400,
766,
294,
2115,
295,
264,
2899,
1242,
51020,
51020,
456,
13,
400,
370,
286,
519,
466,
264,
1154,
321,
362,
281,
5039,
281,
483,
456,
307,
437,
366,
439,
295,
264,
4356,
51268,
51268,
2740,
321,
393,
5039,
2051,
264,
636,
30,
1012,
360,
321,
483,
300,
3603,
22830,
15640,
11,
370,
300,
321,
1322,
264,
51488,
51488,
6896,
293,
264,
5245,
293,
1203,
1646,
281,
312,
1075,
281,
4239,
300,
733,
295,
1752,
51692,
51692
]
},
{
"avg_logprob": -0.06850418677696815,
"compression_ratio": 1.7304964539007093,
"end": 191.52,
"id": 29,
"no_speech_prob": 0.00005061955380369909,
"seek": 17840,
"start": 186.32,
"temperature": 0,
"text": " consumer and for ease of use. But it's probably very far off in terms of the technology getting",
"tokens": [
50364,
439,
1051,
294,
1333,
295,
364,
18473,
636,
13,
407,
286,
519,
300,
300,
576,
312,
264,
2603,
1942,
337,
264,
11,
337,
264,
50760,
50760,
9711,
293,
337,
12708,
295,
764,
13,
583,
309,
311,
1391,
588,
1400,
766,
294,
2115,
295,
264,
2899,
1242,
51020,
51020,
456,
13,
400,
370,
286,
519,
466,
264,
1154,
321,
362,
281,
5039,
281,
483,
456,
307,
437,
366,
439,
295,
264,
4356,
51268,
51268,
2740,
321,
393,
5039,
2051,
264,
636,
30,
1012,
360,
321,
483,
300,
3603,
22830,
15640,
11,
370,
300,
321,
1322,
264,
51488,
51488,
6896,
293,
264,
5245,
293,
1203,
1646,
281,
312,
1075,
281,
4239,
300,
733,
295,
1752,
51692,
51692
]
},
{
"avg_logprob": -0.06850418677696815,
"compression_ratio": 1.7304964539007093,
"end": 196.48000000000002,
"id": 30,
"no_speech_prob": 0.00005061955380369909,
"seek": 17840,
"start": 191.52,
"temperature": 0,
"text": " there. And so I think about the problem we have to solve to get there is what are all of the smaller",
"tokens": [
50364,
439,
1051,
294,
1333,
295,
364,
18473,
636,
13,
407,
286,
519,
300,
300,
576,
312,
264,
2603,
1942,
337,
264,
11,
337,
264,
50760,
50760,
9711,
293,
337,
12708,
295,
764,
13,
583,
309,
311,
1391,
588,
1400,
766,
294,
2115,
295,
264,
2899,
1242,
51020,
51020,
456,
13,
400,
370,
286,
519,
466,
264,
1154,
321,
362,
281,
5039,
281,
483,
456,
307,
437,
366,
439,
295,
264,
4356,
51268,
51268,
2740,
321,
393,
5039,
2051,
264,
636,
30,
1012,
360,
321,
483,
300,
3603,
22830,
15640,
11,
370,
300,
321,
1322,
264,
51488,
51488,
6896,
293,
264,
5245,
293,
1203,
1646,
281,
312,
1075,
281,
4239,
300,
733,
295,
1752,
51692,
51692
]
},
{
"avg_logprob": -0.06850418677696815,
"compression_ratio": 1.7304964539007093,
"end": 200.88,
"id": 31,
"no_speech_prob": 0.00005061955380369909,
"seek": 17840,
"start": 196.48000000000002,
"temperature": 0,
"text": " problems we can solve along the way? How do we get that flywheel spinning, so that we build the",
"tokens": [
50364,
439,
1051,
294,
1333,
295,
364,
18473,
636,
13,
407,
286,
519,
300,
300,
576,
312,
264,
2603,
1942,
337,
264,
11,
337,
264,
50760,
50760,
9711,
293,
337,
12708,
295,
764,
13,
583,
309,
311,
1391,
588,
1400,
766,
294,
2115,
295,
264,
2899,
1242,
51020,
51020,
456,
13,
400,
370,
286,
519,
466,
264,
1154,
321,
362,
281,
5039,
281,
483,
456,
307,
437,
366,
439,
295,
264,
4356,
51268,
51268,
2740,
321,
393,
5039,
2051,
264,
636,
30,
1012,
360,
321,
483,
300,
3603,
22830,
15640,
11,
370,
300,
321,
1322,
264,
51488,
51488,
6896,
293,
264,
5245,
293,
1203,
1646,
281,
312,
1075,
281,
4239,
300,
733,
295,
1752,
51692,
51692
]
},
{
"avg_logprob": -0.06850418677696815,
"compression_ratio": 1.7304964539007093,
"end": 204.96,
"id": 32,
"no_speech_prob": 0.00005061955380369909,
"seek": 17840,
"start": 200.88,
"temperature": 0,
"text": " infrastructure and the models and everything else to be able to deliver that kind of experience",
"tokens": [
50364,
439,
1051,
294,
1333,
295,
364,
18473,
636,
13,
407,
286,
519,
300,
300,
576,
312,
264,
2603,
1942,
337,
264,
11,
337,
264,
50760,
50760,
9711,
293,
337,
12708,
295,
764,
13,
583,
309,
311,
1391,
588,
1400,
766,
294,
2115,
295,
264,
2899,
1242,
51020,
51020,
456,
13,
400,
370,
286,
519,
466,
264,
1154,
321,
362,
281,
5039,
281,
483,
456,
307,
437,
366,
439,
295,
264,
4356,
51268,
51268,
2740,
321,
393,
5039,
2051,
264,
636,
30,
1012,
360,
321,
483,
300,
3603,
22830,
15640,
11,
370,
300,
321,
1322,
264,
51488,
51488,
6896,
293,
264,
5245,
293,
1203,
1646,
281,
312,
1075,
281,
4239,
300,
733,
295,
1752,
51692,
51692
]
},
{
"avg_logprob": -0.11551795334651552,
"compression_ratio": 1.7669172932330828,
"end": 213.68,
"id": 33,
"no_speech_prob": 0.000004565821200230857,
"seek": 20496,
"start": 204.96,
"temperature": 0,
"text": " eventually? Yeah, I mean, one of the companies echoing, I think Jason's point, like one of the",
"tokens": [
50364,
4728,
30,
865,
11,
286,
914,
11,
472,
295,
264,
3431,
14300,
278,
11,
286,
519,
11181,
311,
935,
11,
411,
472,
295,
264,
50800,
50800,
3431,
300,
286,
390,
5665,
588,
2919,
466,
11,
411,
1732,
924,
2057,
11,
420,
746,
11,
51000,
51000,
390,
341,
588,
1627,
7553,
2237,
1219,
1156,
17638,
7348,
13,
400,
641,
1379,
935,
390,
411,
11,
322,
428,
13307,
11,
51304,
51304,
436,
576,
2837,
439,
295,
613,
733,
295,
411,
2973,
1412,
2793,
300,
11181,
390,
1417,
466,
13,
51564,
51564,
400,
294,
641,
13795,
38750,
1680,
10723,
11,
286,
519,
322,
3233,
11,
436,
632,
472,
598,
14917,
1011,
264,
2593,
281,
702,
51832,
51832
]
},
{
"avg_logprob": -0.11551795334651552,
"compression_ratio": 1.7669172932330828,
"end": 217.68,
"id": 34,
"no_speech_prob": 0.000004565821200230857,
"seek": 20496,
"start": 213.68,
"temperature": 0,
"text": " companies that I was personally very excited about, like five years ago, or something,",
"tokens": [
50364,
4728,
30,
865,
11,
286,
914,
11,
472,
295,
264,
3431,
14300,
278,
11,
286,
519,
11181,
311,
935,
11,
411,
472,
295,
264,
50800,
50800,
3431,
300,
286,
390,
5665,
588,
2919,
466,
11,
411,
1732,
924,
2057,
11,
420,
746,
11,
51000,
51000,
390,
341,
588,
1627,
7553,
2237,
1219,
1156,
17638,
7348,
13,
400,
641,
1379,
935,
390,
411,
11,
322,
428,
13307,
11,
51304,
51304,
436,
576,
2837,
439,
295,
613,
733,
295,
411,
2973,
1412,
2793,
300,
11181,
390,
1417,
466,
13,
51564,
51564,
400,
294,
641,
13795,
38750,
1680,
10723,
11,
286,
519,
322,
3233,
11,
436,
632,
472,
598,
14917,
1011,
264,
2593,
281,
702,
51832,
51832
]
},
{
"avg_logprob": -0.11551795334651552,
"compression_ratio": 1.7669172932330828,
"end": 223.76000000000002,
"id": 35,
"no_speech_prob": 0.000004565821200230857,
"seek": 20496,
"start": 217.68,
"temperature": 0,
"text": " was this very cool tech company called Unifi ID. And their whole point was like, on your smartphone,",
"tokens": [
50364,
4728,
30,
865,
11,
286,
914,
11,
472,
295,
264,
3431,
14300,
278,
11,
286,
519,
11181,
311,
935,
11,
411,
472,
295,
264,
50800,
50800,
3431,
300,
286,
390,
5665,
588,
2919,
466,
11,
411,
1732,
924,
2057,
11,
420,
746,
11,
51000,
51000,
390,
341,
588,
1627,
7553,
2237,
1219,
1156,
17638,
7348,
13,
400,
641,
1379,
935,
390,
411,
11,
322,
428,
13307,
11,
51304,
51304,
436,
576,
2837,
439,
295,
613,
733,
295,
411,
2973,
1412,
2793,
300,
11181,
390,
1417,
466,
13,
51564,
51564,
400,
294,
641,
13795,
38750,
1680,
10723,
11,
286,
519,
322,
3233,
11,
436,
632,
472,
598,
14917,
1011,
264,
2593,
281,
702,
51832,
51832
]
},
{
"avg_logprob": -0.11551795334651552,
"compression_ratio": 1.7669172932330828,
"end": 228.96,
"id": 36,
"no_speech_prob": 0.000004565821200230857,
"seek": 20496,
"start": 223.76000000000002,
"temperature": 0,
"text": " they would track all of these kind of like personal data points that Jason was talking about.",
"tokens": [
50364,
4728,
30,
865,
11,
286,
914,
11,
472,
295,
264,
3431,
14300,
278,
11,
286,
519,
11181,
311,
935,
11,
411,
472,
295,
264,
50800,
50800,
3431,
300,
286,
390,
5665,
588,
2919,
466,
11,
411,
1732,
924,
2057,
11,
420,
746,
11,
51000,
51000,
390,
341,
588,
1627,
7553,
2237,
1219,
1156,
17638,
7348,
13,
400,
641,
1379,
935,
390,
411,
11,
322,
428,
13307,
11,
51304,
51304,
436,
576,
2837,
439,
295,
613,
733,
295,
411,
2973,
1412,
2793,
300,
11181,
390,
1417,
466,
13,
51564,
51564,
400,
294,
641,
13795,
38750,
1680,
10723,
11,
286,
519,
322,
3233,
11,
436,
632,
472,
598,
14917,
1011,
264,
2593,
281,
702,
51832,
51832
]
},
{
"avg_logprob": -0.11551795334651552,
"compression_ratio": 1.7669172932330828,
"end": 234.32,
"id": 37,
"no_speech_prob": 0.000004565821200230857,
"seek": 20496,
"start": 228.96,
"temperature": 0,
"text": " And in their TechCrunch demo, I think on stage, they had one co founder hand the phone to his",
"tokens": [
50364,
4728,
30,
865,
11,
286,
914,
11,
472,
295,
264,
3431,
14300,
278,
11,
286,
519,
11181,
311,
935,
11,
411,
472,
295,
264,
50800,
50800,
3431,
300,
286,
390,
5665,
588,
2919,
466,
11,
411,
1732,
924,
2057,
11,
420,
746,
11,
51000,
51000,
390,
341,
588,
1627,
7553,
2237,
1219,
1156,
17638,
7348,
13,
400,
641,
1379,
935,
390,
411,
11,
322,
428,
13307,
11,
51304,
51304,
436,
576,
2837,
439,
295,
613,
733,
295,
411,
2973,
1412,
2793,
300,
11181,
390,
1417,
466,
13,
51564,
51564,
400,
294,
641,
13795,
38750,
1680,
10723,
11,
286,
519,
322,
3233,
11,
436,
632,
472,
598,
14917,
1011,
264,
2593,
281,
702,
51832,
51832
]
},
{
"avg_logprob": -0.0777325756781924,
"compression_ratio": 1.7047970479704797,
"end": 238.88,
"id": 38,
"no_speech_prob": 0.00007966578414198011,
"seek": 23432,
"start": 234.32,
"temperature": 0,
"text": " other co founder, the other co founder walked around with the phone for like a couple seconds,",
"tokens": [
50364,
661,
598,
14917,
11,
264,
661,
598,
14917,
7628,
926,
365,
264,
2593,
337,
411,
257,
1916,
3949,
11,
50592,
50592,
3031,
281,
3565,
294,
11,
293,
309,
848,
15470,
538,
1156,
17638,
7348,
11,
411,
4448,
428,
8539,
307,
406,
264,
912,
382,
50860,
50860,
300,
295,
264,
3380,
7289,
11,
597,
307,
1687,
11,
1687,
10654,
13,
9210,
264,
3430,
456,
307,
550,
51216,
51216,
1242,
264,
5245,
281,
1190,
2329,
379,
11,
570,
264,
8535,
550,
307,
291,
26151,
257,
7154,
11,
51524,
51524,
597,
575,
439,
613,
411,
3228,
29470,
5245,
12187,
322,
309,
13,
400,
264,
7154,
575,
281,
751,
281,
264,
5021,
11,
51772,
51772
]
},
{
"avg_logprob": -0.0777325756781924,
"compression_ratio": 1.7047970479704797,
"end": 244.23999999999998,
"id": 39,
"no_speech_prob": 0.00007966578414198011,
"seek": 23432,
"start": 238.88,
"temperature": 0,
"text": " tried to log in, and it said blocked by Unifi ID, like clearly your gate is not the same as",
"tokens": [
50364,
661,
598,
14917,
11,
264,
661,
598,
14917,
7628,
926,
365,
264,
2593,
337,
411,
257,
1916,
3949,
11,
50592,
50592,
3031,
281,
3565,
294,
11,
293,
309,
848,
15470,
538,
1156,
17638,
7348,
11,
411,
4448,
428,
8539,
307,
406,
264,
912,
382,
50860,
50860,
300,
295,
264,
3380,
7289,
11,
597,
307,
1687,
11,
1687,
10654,
13,
9210,
264,
3430,
456,
307,
550,
51216,
51216,
1242,
264,
5245,
281,
1190,
2329,
379,
11,
570,
264,
8535,
550,
307,
291,
26151,
257,
7154,
11,
51524,
51524,
597,
575,
439,
613,
411,
3228,
29470,
5245,
12187,
322,
309,
13,
400,
264,
7154,
575,
281,
751,
281,
264,
5021,
11,
51772,
51772
]
},
{
"avg_logprob": -0.0777325756781924,
"compression_ratio": 1.7047970479704797,
"end": 251.35999999999999,
"id": 40,
"no_speech_prob": 0.00007966578414198011,
"seek": 23432,
"start": 244.23999999999998,
"temperature": 0,
"text": " that of the original owner, which is super, super neat. Probably the challenge there is then",
"tokens": [
50364,
661,
598,
14917,
11,
264,
661,
598,
14917,
7628,
926,
365,
264,
2593,
337,
411,
257,
1916,
3949,
11,
50592,
50592,
3031,
281,
3565,
294,
11,
293,
309,
848,
15470,
538,
1156,
17638,
7348,
11,
411,
4448,
428,
8539,
307,
406,
264,
912,
382,
50860,
50860,
300,
295,
264,
3380,
7289,
11,
597,
307,
1687,
11,
1687,
10654,
13,
9210,
264,
3430,
456,
307,
550,
51216,
51216,
1242,
264,
5245,
281,
1190,
2329,
379,
11,
570,
264,
8535,
550,
307,
291,
26151,
257,
7154,
11,
51524,
51524,
597,
575,
439,
613,
411,
3228,
29470,
5245,
12187,
322,
309,
13,
400,
264,
7154,
575,
281,
751,
281,
264,
5021,
11,
51772,
51772
]
},
{
"avg_logprob": -0.0777325756781924,
"compression_ratio": 1.7047970479704797,
"end": 257.52,
"id": 41,
"no_speech_prob": 0.00007966578414198011,
"seek": 23432,
"start": 251.35999999999999,
"temperature": 0,
"text": " getting the models to run vocally, because the alternative then is you ping a server,",
"tokens": [
50364,
661,
598,
14917,
11,
264,
661,
598,
14917,
7628,
926,
365,
264,
2593,
337,
411,
257,
1916,
3949,
11,
50592,
50592,
3031,
281,
3565,
294,
11,
293,
309,
848,
15470,
538,
1156,
17638,
7348,
11,
411,
4448,
428,
8539,
307,
406,
264,
912,
382,
50860,
50860,
300,
295,
264,
3380,
7289,
11,
597,
307,
1687,
11,
1687,
10654,
13,
9210,
264,
3430,
456,
307,
550,
51216,
51216,
1242,
264,
5245,
281,
1190,
2329,
379,
11,
570,
264,
8535,
550,
307,
291,
26151,
257,
7154,
11,
51524,
51524,
597,
575,
439,
613,
411,
3228,
29470,
5245,
12187,
322,
309,
13,
400,
264,
7154,
575,
281,
751,
281,
264,
5021,
11,
51772,
51772
]
},
{
"avg_logprob": -0.0777325756781924,
"compression_ratio": 1.7047970479704797,
"end": 262.48,
"id": 42,
"no_speech_prob": 0.00007966578414198011,
"seek": 23432,
"start": 257.52,
"temperature": 0,
"text": " which has all these like biometric models stored on it. And the server has to talk to the chain,",
"tokens": [
50364,
661,
598,
14917,
11,
264,
661,
598,
14917,
7628,
926,
365,
264,
2593,
337,
411,
257,
1916,
3949,
11,
50592,
50592,
3031,
281,
3565,
294,
11,
293,
309,
848,
15470,
538,
1156,
17638,
7348,
11,
411,
4448,
428,
8539,
307,
406,
264,
912,
382,
50860,
50860,
300,
295,
264,
3380,
7289,
11,
597,
307,
1687,
11,
1687,
10654,
13,
9210,
264,
3430,
456,
307,
550,
51216,
51216,
1242,
264,
5245,
281,
1190,
2329,
379,
11,
570,
264,
8535,
550,
307,
291,
26151,
257,
7154,
11,
51524,
51524,
597,
575,
439,
613,
411,
3228,
29470,
5245,
12187,
322,
309,
13,
400,
264,
7154,
575,
281,
751,
281,
264,
5021,
11,
51772,
51772
]
},
{
"avg_logprob": -0.09056083779585988,
"compression_ratio": 1.7551867219917012,
"end": 267.04,
"id": 43,
"no_speech_prob": 0.000046110213588690385,
"seek": 26248,
"start": 262.48,
"temperature": 0,
"text": " and you have to talk to the server. And there's all this indirection. So the ideal here would",
"tokens": [
50364,
293,
291,
362,
281,
751,
281,
264,
7154,
13,
400,
456,
311,
439,
341,
1016,
621,
882,
13,
407,
264,
7157,
510,
576,
50592,
50592,
445,
312,
291,
362,
439,
264,
5245,
322,
2593,
11,
293,
291,
393,
1190,
264,
5245,
11,
291,
393,
1190,
8177,
82,
337,
50824,
50824,
264,
5245,
13,
400,
24202,
365,
633,
14425,
307,
1936,
257,
8177,
295,
1566,
411,
11,
2086,
11,
341,
390,
51116,
51116,
767,
2279,
538,
385,
13,
400,
538,
264,
636,
11,
510,
311,
437,
286,
528,
281,
360,
11,
411,
572,
544,
643,
337,
257,
4054,
2141,
11,
51360,
51360,
572,
544,
643,
337,
604,
295,
300,
13,
865,
13,
51492
]
},
{
"avg_logprob": -0.09056083779585988,
"compression_ratio": 1.7551867219917012,
"end": 271.68,
"id": 44,
"no_speech_prob": 0.000046110213588690385,
"seek": 26248,
"start": 267.04,
"temperature": 0,
"text": " just be you have all the models on phone, and you can run the models, you can run proofs for",
"tokens": [
50364,
293,
291,
362,
281,
751,
281,
264,
7154,
13,
400,
456,
311,
439,
341,
1016,
621,
882,
13,
407,
264,
7157,
510,
576,
50592,
50592,
445,
312,
291,
362,
439,
264,
5245,
322,
2593,
11,
293,
291,
393,
1190,
264,
5245,
11,
291,
393,
1190,
8177,
82,
337,
50824,
50824,
264,
5245,
13,
400,
24202,
365,
633,
14425,
307,
1936,
257,
8177,
295,
1566,
411,
11,
2086,
11,
341,
390,
51116,
51116,
767,
2279,
538,
385,
13,
400,
538,
264,
636,
11,
510,
311,
437,
286,
528,
281,
360,
11,
411,
572,
544,
643,
337,
257,
4054,
2141,
11,
51360,
51360,
572,
544,
643,
337,
604,
295,
300,
13,
865,
13,
51492
]
},
{
"avg_logprob": -0.09056083779585988,
"compression_ratio": 1.7551867219917012,
"end": 277.52000000000004,
"id": 45,
"no_speech_prob": 0.000046110213588690385,
"seek": 26248,
"start": 271.68,
"temperature": 0,
"text": " the models. And accompanied with every transaction is basically a proof of saying like, yes, this was",
"tokens": [
50364,
293,
291,
362,
281,
751,
281,
264,
7154,
13,
400,
456,
311,
439,
341,
1016,
621,
882,
13,
407,
264,
7157,
510,
576,
50592,
50592,
445,
312,
291,
362,
439,
264,
5245,
322,
2593,
11,
293,
291,
393,
1190,
264,
5245,
11,
291,
393,
1190,
8177,
82,
337,
50824,
50824,
264,
5245,
13,
400,
24202,
365,
633,
14425,
307,
1936,
257,
8177,
295,
1566,
411,
11,
2086,
11,
341,
390,
51116,
51116,
767,
2279,
538,
385,
13,
400,
538,
264,
636,
11,
510,
311,
437,
286,
528,
281,
360,
11,
411,
572,
544,
643,
337,
257,
4054,
2141,
11,
51360,
51360,
572,
544,
643,
337,
604,
295,
300,
13,
865,
13,
51492
]
},
{
"avg_logprob": -0.09056083779585988,
"compression_ratio": 1.7551867219917012,
"end": 282.40000000000003,
"id": 46,
"no_speech_prob": 0.000046110213588690385,
"seek": 26248,
"start": 277.52000000000004,
"temperature": 0,
"text": " actually sent by me. And by the way, here's what I want to do, like no more need for a secret key,",
"tokens": [
50364,
293,
291,
362,
281,
751,
281,
264,
7154,
13,
400,
456,
311,
439,
341,
1016,
621,
882,
13,
407,
264,
7157,
510,
576,
50592,
50592,
445,
312,
291,
362,
439,
264,
5245,
322,
2593,
11,
293,
291,
393,
1190,
264,
5245,
11,
291,
393,
1190,
8177,
82,
337,
50824,
50824,
264,
5245,
13,
400,
24202,
365,
633,
14425,
307,
1936,
257,
8177,
295,
1566,
411,
11,
2086,
11,
341,
390,
51116,
51116,
767,
2279,
538,
385,
13,
400,
538,
264,
636,
11,
510,
311,
437,
286,
528,
281,
360,
11,
411,
572,
544,
643,
337,
257,
4054,
2141,
11,
51360,
51360,
572,
544,
643,
337,
604,
295,
300,
13,
865,
13,
51492
]
},
{
"avg_logprob": -0.331805402582342,
"compression_ratio": 1.3,
"end": 298,
"id": 47,
"no_speech_prob": 0.00032447325065732,
"seek": 28240,
"start": 282.4,
"temperature": 0,
"text": " no more need for any of that. Yeah. Corin, folks, you want to add something to that?",
"tokens": [
50364,
572,
544,
643,
337,
604,
295,
300,
13,
865,
13,
3925,
259,
11,
4024,
11,
291,
528,
281,
909,
746,
281,
300,
30,
51144,
51420,
509,
458,
437,
11,
3480,
493,
264,
15471,
11,
6792,
5629,
294,
264,
5081,
11,
51756,
51808
]
},
{
"avg_logprob": -0.331805402582342,
"compression_ratio": 1.3,
"end": 310.23999999999995,
"id": 48,
"no_speech_prob": 0.00032447325065732,
"seek": 28240,
"start": 303.52,
"temperature": 0,
"text": " You know what, following up the Jonathan, latest statement in the chat,",
"tokens": [
50364,
572,
544,
643,
337,
604,
295,
300,
13,
865,
13,
3925,
259,
11,
4024,
11,
291,
528,
281,
909,
746,
281,
300,
30,
51144,
51420,
509,
458,
437,
11,
3480,
493,
264,
15471,
11,
6792,
5629,
294,
264,
5081,
11,
51756,
51808
]
},
{
"avg_logprob": -0.2941874663035075,
"compression_ratio": 1.4484536082474226,
"end": 316.96000000000004,
"id": 49,
"no_speech_prob": 0.00005220569801167585,
"seek": 31024,
"start": 310.24,
"temperature": 0,
"text": " we'd love to know if there is some, but this is more for machine learning than ZK, a way to",
"tokens": [
50364,
321,
1116,
959,
281,
458,
498,
456,
307,
512,
11,
457,
341,
307,
544,
337,
3479,
2539,
813,
1176,
42,
11,
257,
636,
281,
50700,
50772,
1223,
498,
732,
420,
819,
5794,
23693,
366,
4077,
281,
264,
912,
954,
11,
291,
458,
11,
293,
51232,
51264,
1228,
1176,
42,
12683,
420,
341,
733,
295,
2890,
281,
2893,
512,
1333,
295,
383,
17624,
7335,
337,
11,
2318,
3938,
11,
51728,
51728
]
},
{
"avg_logprob": -0.2941874663035075,
"compression_ratio": 1.4484536082474226,
"end": 327.6,
"id": 50,
"no_speech_prob": 0.00005220569801167585,
"seek": 31024,
"start": 318.40000000000003,
"temperature": 0,
"text": " understand if two or different Twitter profiles are related to the same person, you know, and",
"tokens": [
50364,
321,
1116,
959,
281,
458,
498,
456,
307,
512,
11,
457,
341,
307,
544,
337,
3479,
2539,
813,
1176,
42,
11,
257,
636,
281,
50700,
50772,
1223,
498,
732,
420,
819,
5794,
23693,
366,
4077,
281,
264,
912,
954,
11,
291,
458,
11,
293,
51232,
51264,
1228,
1176,
42,
12683,
420,
341,
733,
295,
2890,
281,
2893,
512,
1333,
295,
383,
17624,
7335,
337,
11,
2318,
3938,
11,
51728,
51728
]
},
{
"avg_logprob": -0.2941874663035075,
"compression_ratio": 1.4484536082474226,
"end": 337.52,
"id": 51,
"no_speech_prob": 0.00005220569801167585,
"seek": 31024,
"start": 328.24,
"temperature": 0,
"text": " using ZKML or this kind of mix to provide some sort of CBL resistance for, especially recently,",
"tokens": [
50364,
321,
1116,
959,
281,
458,
498,
456,
307,
512,
11,
457,
341,
307,
544,
337,
3479,
2539,
813,
1176,
42,
11,
257,
636,
281,
50700,
50772,
1223,
498,
732,
420,
819,
5794,
23693,
366,
4077,
281,
264,
912,
954,
11,
291,
458,
11,
293,
51232,
51264,
1228,
1176,
42,
12683,
420,
341,
733,
295,
2890,
281,
2893,
512,
1333,
295,
383,
17624,
7335,
337,
11,
2318,
3938,
11,
51728,
51728
]
},
{
"avg_logprob": -0.21151840940434882,
"compression_ratio": 1.6740088105726871,
"end": 345.12,
"id": 52,
"no_speech_prob": 0.00008216579590225592,
"seek": 33752,
"start": 337.52,
"temperature": 0,
"text": " there was a discussion of, I mean, we already knew that someone is doing that for, you know,",
"tokens": [
50364,
456,
390,
257,
5017,
295,
11,
286,
914,
11,
321,
1217,
2586,
300,
1580,
307,
884,
300,
337,
11,
291,
458,
11,
50744,
50872,
884,
2531,
8122,
337,
257,
1271,
49715,
13,
509,
4411,
1391,
365,
1156,
271,
86,
569,
562,
291,
764,
36116,
2696,
51204,
51204,
300,
1936,
291,
4774,
867,
11438,
11,
291,
458,
11,
293,
370,
1580,
307,
1382,
281,
9987,
1125,
309,
13,
51504,
51504,
407,
286,
519,
300,
337,
11,
291,
458,
11,
456,
366,
867,
15257,
411,
8177,
295,
10243,
300,
360,
257,
960,
51776,
51776
]
},
{
"avg_logprob": -0.21151840940434882,
"compression_ratio": 1.6740088105726871,
"end": 354.32,
"id": 53,
"no_speech_prob": 0.00008216579590225592,
"seek": 33752,
"start": 347.68,
"temperature": 0,
"text": " doing similar attacks for airdrops. You discover probably with Uniswap when you use burner account",
"tokens": [
50364,
456,
390,
257,
5017,
295,
11,
286,
914,
11,
321,
1217,
2586,
300,
1580,
307,
884,
300,
337,
11,
291,
458,
11,
50744,
50872,
884,
2531,
8122,
337,
257,
1271,
49715,
13,
509,
4411,
1391,
365,
1156,
271,
86,
569,
562,
291,
764,
36116,
2696,
51204,
51204,
300,
1936,
291,
4774,
867,
11438,
11,
291,
458,
11,
293,
370,
1580,
307,
1382,
281,
9987,
1125,
309,
13,
51504,
51504,
407,
286,
519,
300,
337,
11,
291,
458,
11,
456,
366,
867,
15257,
411,
8177,
295,
10243,
300,
360,
257,
960,
51776,
51776
]
},
{
"avg_logprob": -0.21151840940434882,
"compression_ratio": 1.6740088105726871,
"end": 360.32,
"id": 54,
"no_speech_prob": 0.00008216579590225592,
"seek": 33752,
"start": 354.32,
"temperature": 0,
"text": " that basically you receive many drops, you know, and so someone is trying to industrialize it.",
"tokens": [
50364,
456,
390,
257,
5017,
295,
11,
286,
914,
11,
321,
1217,
2586,
300,
1580,
307,
884,
300,
337,
11,
291,
458,
11,
50744,
50872,
884,
2531,
8122,
337,
257,
1271,
49715,
13,
509,
4411,
1391,
365,
1156,
271,
86,
569,
562,
291,
764,
36116,
2696,
51204,
51204,
300,
1936,
291,
4774,
867,
11438,
11,
291,
458,
11,
293,
370,
1580,
307,
1382,
281,
9987,
1125,
309,
13,
51504,
51504,
407,
286,
519,
300,
337,
11,
291,
458,
11,
456,
366,
867,
15257,
411,
8177,
295,
10243,
300,
360,
257,
960,
51776,
51776
]
},
{
"avg_logprob": -0.21151840940434882,
"compression_ratio": 1.6740088105726871,
"end": 365.76,
"id": 55,
"no_speech_prob": 0.00008216579590225592,
"seek": 33752,
"start": 360.32,
"temperature": 0,
"text": " So I think that for, you know, there are many attempts like proof of humanity that do a video",
"tokens": [
50364,
456,
390,
257,
5017,
295,
11,
286,
914,
11,
321,
1217,
2586,
300,
1580,
307,
884,
300,
337,
11,
291,
458,
11,
50744,
50872,
884,
2531,
8122,
337,
257,
1271,
49715,
13,
509,
4411,
1391,
365,
1156,
271,
86,
569,
562,
291,
764,
36116,
2696,
51204,
51204,
300,
1936,
291,
4774,
867,
11438,
11,
291,
458,
11,
293,
370,
1580,
307,
1382,
281,
9987,
1125,
309,
13,
51504,
51504,
407,
286,
519,
300,
337,
11,
291,
458,
11,
456,
366,
867,
15257,
411,
8177,
295,
10243,
300,
360,
257,
960,
51776,
51776
]
},
{
"avg_logprob": -0.21213083556204132,
"compression_ratio": 1.5081967213114753,
"end": 371.44,
"id": 56,
"no_speech_prob": 0.0009516674326732755,
"seek": 36576,
"start": 365.76,
"temperature": 0,
"text": " recording with your wallet address, you've seen a transaction chain, you know, there are many",
"tokens": [
50364,
6613,
365,
428,
16599,
2985,
11,
291,
600,
1612,
257,
14425,
5021,
11,
291,
458,
11,
456,
366,
867,
50648,
50648,
11587,
300,
321,
393,
7958,
466,
264,
3825,
17695,
293,
867,
11,
867,
2357,
11,
1392,
11,
51004,
51064,
300,
1580,
307,
4099,
766,
13,
583,
1338,
11,
264,
1168,
337,
385,
307,
294,
2674,
11,
718,
311,
584,
11,
7228,
51364,
51364
]
},
{
"avg_logprob": -0.21213083556204132,
"compression_ratio": 1.5081967213114753,
"end": 378.56,
"id": 57,
"no_speech_prob": 0.0009516674326732755,
"seek": 36576,
"start": 371.44,
"temperature": 0,
"text": " approaches that we can debate about the security assumptions and many, many others, okay,",
"tokens": [
50364,
6613,
365,
428,
16599,
2985,
11,
291,
600,
1612,
257,
14425,
5021,
11,
291,
458,
11,
456,
366,
867,
50648,
50648,
11587,
300,
321,
393,
7958,
466,
264,
3825,
17695,
293,
867,
11,
867,
2357,
11,
1392,
11,
51004,
51064,
300,
1580,
307,
4099,
766,
13,
583,
1338,
11,
264,
1168,
337,
385,
307,
294,
2674,
11,
718,
311,
584,
11,
7228,
51364,
51364
]
},
{
"avg_logprob": -0.21213083556204132,
"compression_ratio": 1.5081967213114753,
"end": 385.76,
"id": 58,
"no_speech_prob": 0.0009516674326732755,
"seek": 36576,
"start": 379.76,
"temperature": 0,
"text": " that someone is showing off. But yeah, the question for me is in general, let's say, despite",
"tokens": [
50364,
6613,
365,
428,
16599,
2985,
11,
291,
600,
1612,
257,
14425,
5021,
11,
291,
458,
11,
456,
366,
867,
50648,
50648,
11587,
300,
321,
393,
7958,
466,
264,
3825,
17695,
293,
867,
11,
867,
2357,
11,
1392,
11,
51004,
51064,
300,
1580,
307,
4099,
766,
13,
583,
1338,
11,
264,
1168,
337,
385,
307,
294,
2674,
11,
718,
311,
584,
11,
7228,
51364,
51364
]
},
{
"avg_logprob": -0.2381147971520057,
"compression_ratio": 1.5027624309392265,
"end": 395.68,
"id": 59,
"no_speech_prob": 0.00006399819540092722,
"seek": 38576,
"start": 385.76,
"temperature": 0,
"text": " the specific protocol, whatever, taking an approach of working backwards to see which could be the",
"tokens": [
50364,
264,
2685,
10336,
11,
2035,
11,
1940,
364,
3109,
295,
1364,
12204,
281,
536,
597,
727,
312,
264,
50860,
50884,
764,
1389,
11,
577,
360,
291,
519,
300,
321,
393,
3079,
341,
2899,
281,
729,
721,
30,
51124,
51248,
286,
914,
11,
341,
307,
1936,
437,
321,
434,
884,
13,
1485,
295,
264,
1152,
2740,
510,
307,
25775,
309,
646,
281,
51604,
51604
]
},
{
"avg_logprob": -0.2381147971520057,
"compression_ratio": 1.5027624309392265,
"end": 400.96,
"id": 60,
"no_speech_prob": 0.00006399819540092722,
"seek": 38576,
"start": 396.15999999999997,
"temperature": 0,
"text": " use case, how do you think that we can apply this technology to those things?",
"tokens": [
50364,
264,
2685,
10336,
11,
2035,
11,
1940,
364,
3109,
295,
1364,
12204,
281,
536,
597,
727,
312,
264,
50860,
50884,
764,
1389,
11,
577,
360,
291,
519,
300,
321,
393,
3079,
341,
2899,
281,
729,
721,
30,
51124,
51248,
286,
914,
11,
341,
307,
1936,
437,
321,
434,
884,
13,
1485,
295,
264,
1152,
2740,
510,
307,
25775,
309,
646,
281,
51604,
51604
]
},
{
"avg_logprob": -0.2381147971520057,
"compression_ratio": 1.5027624309392265,
"end": 410.56,
"id": 61,
"no_speech_prob": 0.00006399819540092722,
"seek": 38576,
"start": 403.44,
"temperature": 0,
"text": " I mean, this is basically what we're doing. One of the hard problems here is linking it back to",
"tokens": [
50364,
264,
2685,
10336,
11,
2035,
11,
1940,
364,
3109,
295,
1364,
12204,
281,
536,
597,
727,
312,
264,
50860,
50884,
764,
1389,
11,
577,
360,
291,
519,
300,
321,
393,
3079,
341,
2899,
281,
729,
721,
30,
51124,
51248,
286,
914,
11,
341,
307,
1936,
437,
321,
434,
884,
13,
1485,
295,
264,
1152,
2740,
510,
307,
25775,
309,
646,
281,
51604,
51604
]
},
{
"avg_logprob": -0.16084403617709292,
"compression_ratio": 1.625,
"end": 417.28000000000003,
"id": 62,
"no_speech_prob": 0.00026054971385747194,
"seek": 41056,
"start": 410.56,
"temperature": 0,
"text": " reality. So getting some sort of trusted, you need some sort of trusted Oracle until a physicist",
"tokens": [
50364,
4103,
13,
407,
1242,
512,
1333,
295,
16034,
11,
291,
643,
512,
1333,
295,
16034,
25654,
1826,
257,
42466,
50700,
50700,
8932,
484,
577,
281,
4445,
413,
3661,
414,
12,
39,
898,
1601,
294,
264,
6064,
295,
3687,
13,
400,
264,
636,
321,
360,
300,
307,
538,
264,
51080,
51080,
16997,
707,
14715,
300,
264,
413,
57,
27377,
445,
4712,
322,
2799,
11,
291,
393,
920,
536,
309,
2261,
796,
322,
257,
51356,
51356,
1463,
13,
407,
1564,
291,
3241,
300,
11,
293,
1564,
291,
483,
670,
300,
47423,
295,
11,
1392,
11,
321,
643,
512,
1254,
295,
51668,
51668
]
},
{
"avg_logprob": -0.16084403617709292,
"compression_ratio": 1.625,
"end": 424.88,
"id": 63,
"no_speech_prob": 0.00026054971385747194,
"seek": 41056,
"start": 417.28000000000003,
"temperature": 0,
"text": " figured out how to implement Diffie-Hellman in the laws of nature. And the way we do that is by the",
"tokens": [
50364,
4103,
13,
407,
1242,
512,
1333,
295,
16034,
11,
291,
643,
512,
1333,
295,
16034,
25654,
1826,
257,
42466,
50700,
50700,
8932,
484,
577,
281,
4445,
413,
3661,
414,
12,
39,
898,
1601,
294,
264,
6064,
295,
3687,
13,
400,
264,
636,
321,
360,
300,
307,
538,
264,
51080,
51080,
16997,
707,
14715,
300,
264,
413,
57,
27377,
445,
4712,
322,
2799,
11,
291,
393,
920,
536,
309,
2261,
796,
322,
257,
51356,
51356,
1463,
13,
407,
1564,
291,
3241,
300,
11,
293,
1564,
291,
483,
670,
300,
47423,
295,
11,
1392,
11,
321,
643,
512,
1254,
295,
51668,
51668
]
},
{
"avg_logprob": -0.16084403617709292,
"compression_ratio": 1.625,
"end": 430.4,
"id": 64,
"no_speech_prob": 0.00026054971385747194,
"seek": 41056,
"start": 424.88,
"temperature": 0,
"text": " shiny little orb that the DZ builder just showed on camera, you can still see it behind him on a",
"tokens": [
50364,
4103,
13,
407,
1242,
512,
1333,
295,
16034,
11,
291,
643,
512,
1333,
295,
16034,
25654,
1826,
257,
42466,
50700,
50700,
8932,
484,
577,
281,
4445,
413,
3661,
414,
12,
39,
898,
1601,
294,
264,
6064,
295,
3687,
13,
400,
264,
636,
321,
360,
300,
307,
538,
264,
51080,
51080,
16997,
707,
14715,
300,
264,
413,
57,
27377,
445,
4712,
322,
2799,
11,
291,
393,
920,
536,
309,
2261,
796,
322,
257,
51356,
51356,
1463,
13,
407,
1564,
291,
3241,
300,
11,
293,
1564,
291,
483,
670,
300,
47423,
295,
11,
1392,
11,
321,
643,
512,
1254,
295,
51668,
51668
]
},
{
"avg_logprob": -0.16084403617709292,
"compression_ratio": 1.625,
"end": 436.64,
"id": 65,
"no_speech_prob": 0.00026054971385747194,
"seek": 41056,
"start": 430.4,
"temperature": 0,
"text": " stand. So once you accept that, and once you get over that hurdle of, okay, we need some form of",
"tokens": [
50364,
4103,
13,
407,
1242,
512,
1333,
295,
16034,
11,
291,
643,
512,
1333,
295,
16034,
25654,
1826,
257,
42466,
50700,
50700,
8932,
484,
577,
281,
4445,
413,
3661,
414,
12,
39,
898,
1601,
294,
264,
6064,
295,
3687,
13,
400,
264,
636,
321,
360,
300,
307,
538,
264,
51080,
51080,
16997,
707,
14715,
300,
264,
413,
57,
27377,
445,
4712,
322,
2799,
11,
291,
393,
920,
536,
309,
2261,
796,
322,
257,
51356,
51356,
1463,
13,
407,
1564,
291,
3241,
300,
11,
293,
1564,
291,
483,
670,
300,
47423,
295,
11,
1392,
11,
321,
643,
512,
1254,
295,
51668,
51668
]
},
{
"avg_logprob": -0.18173195213399906,
"compression_ratio": 1.6118143459915613,
"end": 442.15999999999997,
"id": 66,
"no_speech_prob": 0.00023396719188895077,
"seek": 43664,
"start": 436.64,
"temperature": 0,
"text": " trusted hardware to hook into reality, then you move the problem to okay, now we have authentic",
"tokens": [
50364,
16034,
8837,
281,
6328,
666,
4103,
11,
550,
291,
1286,
264,
1154,
281,
1392,
11,
586,
321,
362,
12466,
50640,
50640,
3228,
29470,
1412,
13,
400,
550,
291,
393,
360,
439,
7527,
295,
1880,
721,
1228,
383,
42,
12683,
281,
1399,
300,
1412,
50956,
51012,
294,
257,
636,
300,
2709,
291,
2698,
26976,
293,
307,
32870,
293,
3361,
1832,
293,
439,
729,
1481,
51284,
51284,
7221,
300,
321,
528,
13,
400,
341,
307,
767,
577,
286,
658,
281,
264,
2519,
295,
411,
11,
1176,
42,
12683,
294,
264,
700,
51648,
51648
]
},
{
"avg_logprob": -0.18173195213399906,
"compression_ratio": 1.6118143459915613,
"end": 448.47999999999996,
"id": 67,
"no_speech_prob": 0.00023396719188895077,
"seek": 43664,
"start": 442.15999999999997,
"temperature": 0,
"text": " biometric data. And then you can do all sorts of interesting things using CKML to process that data",
"tokens": [
50364,
16034,
8837,
281,
6328,
666,
4103,
11,
550,
291,
1286,
264,
1154,
281,
1392,
11,
586,
321,
362,
12466,
50640,
50640,
3228,
29470,
1412,
13,
400,
550,
291,
393,
360,
439,
7527,
295,
1880,
721,
1228,
383,
42,
12683,
281,
1399,
300,
1412,
50956,
51012,
294,
257,
636,
300,
2709,
291,
2698,
26976,
293,
307,
32870,
293,
3361,
1832,
293,
439,
729,
1481,
51284,
51284,
7221,
300,
321,
528,
13,
400,
341,
307,
767,
577,
286,
658,
281,
264,
2519,
295,
411,
11,
1176,
42,
12683,
294,
264,
700,
51648,
51648
]
},
{
"avg_logprob": -0.18173195213399906,
"compression_ratio": 1.6118143459915613,
"end": 455.03999999999996,
"id": 68,
"no_speech_prob": 0.00023396719188895077,
"seek": 43664,
"start": 449.59999999999997,
"temperature": 0,
"text": " in a way that gives you self custody and is decentralized and trustless and all those nice",
"tokens": [
50364,
16034,
8837,
281,
6328,
666,
4103,
11,
550,
291,
1286,
264,
1154,
281,
1392,
11,
586,
321,
362,
12466,
50640,
50640,
3228,
29470,
1412,
13,
400,
550,
291,
393,
360,
439,
7527,
295,
1880,
721,
1228,
383,
42,
12683,
281,
1399,
300,
1412,
50956,
51012,
294,
257,
636,
300,
2709,
291,
2698,
26976,
293,
307,
32870,
293,
3361,
1832,
293,
439,
729,
1481,
51284,
51284,
7221,
300,
321,
528,
13,
400,
341,
307,
767,
577,
286,
658,
281,
264,
2519,
295,
411,
11,
1176,
42,
12683,
294,
264,
700,
51648,
51648
]
},
{
"avg_logprob": -0.18173195213399906,
"compression_ratio": 1.6118143459915613,
"end": 462.32,
"id": 69,
"no_speech_prob": 0.00023396719188895077,
"seek": 43664,
"start": 455.03999999999996,
"temperature": 0,
"text": " properties that we want. And this is actually how I got to the field of like, ZKML in the first",
"tokens": [
50364,
16034,
8837,
281,
6328,
666,
4103,
11,
550,
291,
1286,
264,
1154,
281,
1392,
11,
586,
321,
362,
12466,
50640,
50640,
3228,
29470,
1412,
13,
400,
550,
291,
393,
360,
439,
7527,
295,
1880,
721,
1228,
383,
42,
12683,
281,
1399,
300,
1412,
50956,
51012,
294,
257,
636,
300,
2709,
291,
2698,
26976,
293,
307,
32870,
293,
3361,
1832,
293,
439,
729,
1481,
51284,
51284,
7221,
300,
321,
528,
13,
400,
341,
307,
767,
577,
286,
658,
281,
264,
2519,
295,
411,
11,
1176,
42,
12683,
294,
264,
700,
51648,
51648
]
},
{
"avg_logprob": -0.16269845763842264,
"compression_ratio": 1.5673469387755101,
"end": 467.76,
"id": 70,
"no_speech_prob": 0.00017393362941220403,
"seek": 46232,
"start": 462.32,
"temperature": 0,
"text": " place, it was basically right now the orb does the biometric processing and trusted hardware. And my",
"tokens": [
50364,
1081,
11,
309,
390,
1936,
558,
586,
264,
14715,
775,
264,
3228,
29470,
9007,
293,
16034,
8837,
13,
400,
452,
50636,
50636,
1168,
390,
11,
1392,
11,
437,
498,
321,
528,
281,
5623,
264,
2316,
30,
4402,
586,
1518,
643,
281,
352,
646,
281,
341,
50856,
50856,
14715,
30,
400,
321,
393,
5039,
300,
365,
383,
42,
12683,
538,
1936,
445,
13393,
264,
8936,
1412,
293,
884,
439,
264,
51292,
51292,
9007,
294,
264,
4195,
311,
16599,
2564,
11,
597,
39890,
257,
2603,
40176,
2734,
337,
505,
13,
400,
550,
309,
311,
51784,
51784
]
},
{
"avg_logprob": -0.16269845763842264,
"compression_ratio": 1.5673469387755101,
"end": 472.15999999999997,
"id": 71,
"no_speech_prob": 0.00017393362941220403,
"seek": 46232,
"start": 467.76,
"temperature": 0,
"text": " question was, okay, what if we want to update the model? Does now everyone need to go back to this",
"tokens": [
50364,
1081,
11,
309,
390,
1936,
558,
586,
264,
14715,
775,
264,
3228,
29470,
9007,
293,
16034,
8837,
13,
400,
452,
50636,
50636,
1168,
390,
11,
1392,
11,
437,
498,
321,
528,
281,
5623,
264,
2316,
30,
4402,
586,
1518,
643,
281,
352,
646,
281,
341,
50856,
50856,
14715,
30,
400,
321,
393,
5039,
300,
365,
383,
42,
12683,
538,
1936,
445,
13393,
264,
8936,
1412,
293,
884,
439,
264,
51292,
51292,
9007,
294,
264,
4195,
311,
16599,
2564,
11,
597,
39890,
257,
2603,
40176,
2734,
337,
505,
13,
400,
550,
309,
311,
51784,
51784
]
},
{
"avg_logprob": -0.16269845763842264,
"compression_ratio": 1.5673469387755101,
"end": 480.88,
"id": 72,
"no_speech_prob": 0.00017393362941220403,
"seek": 46232,
"start": 472.15999999999997,
"temperature": 0,
"text": " orb? And we can solve that with CKML by basically just signing the raw data and doing all the",
"tokens": [
50364,
1081,
11,
309,
390,
1936,
558,
586,
264,
14715,
775,
264,
3228,
29470,
9007,
293,
16034,
8837,
13,
400,
452,
50636,
50636,
1168,
390,
11,
1392,
11,
437,
498,
321,
528,
281,
5623,
264,
2316,
30,
4402,
586,
1518,
643,
281,
352,
646,
281,
341,
50856,
50856,
14715,
30,
400,
321,
393,
5039,
300,
365,
383,
42,
12683,
538,
1936,
445,
13393,
264,
8936,
1412,
293,
884,
439,
264,
51292,
51292,
9007,
294,
264,
4195,
311,
16599,
2564,
11,
597,
39890,
257,
2603,
40176,
2734,
337,
505,
13,
400,
550,
309,
311,
51784,
51784
]
},
{
"avg_logprob": -0.16269845763842264,
"compression_ratio": 1.5673469387755101,
"end": 490.71999999999997,
"id": 73,
"no_speech_prob": 0.00017393362941220403,
"seek": 46232,
"start": 480.88,
"temperature": 0,
"text": " processing in the user's wallet itself, which solves a huge UX issue for us. And then it's",
"tokens": [
50364,
1081,
11,
309,
390,
1936,
558,
586,
264,
14715,
775,
264,
3228,
29470,
9007,
293,
16034,
8837,
13,
400,
452,
50636,
50636,
1168,
390,
11,
1392,
11,
437,
498,
321,
528,
281,
5623,
264,
2316,
30,
4402,
586,
1518,
643,
281,
352,
646,
281,
341,
50856,
50856,
14715,
30,
400,
321,
393,
5039,
300,
365,
383,
42,
12683,
538,
1936,
445,
13393,
264,
8936,
1412,
293,
884,
439,
264,
51292,
51292,
9007,
294,
264,
4195,
311,
16599,
2564,
11,
597,
39890,
257,
2603,
40176,
2734,
337,
505,
13,
400,
550,
309,
311,
51784,
51784
]
},
{
"avg_logprob": -0.1495452880859375,
"compression_ratio": 1.711111111111111,
"end": 495.52000000000004,
"id": 74,
"no_speech_prob": 0.00004129850276513025,
"seek": 49072,
"start": 490.72,
"temperature": 0,
"text": " turns out that this actually opens a number of other interesting doors. If there is a way for",
"tokens": [
50364,
4523,
484,
300,
341,
767,
9870,
257,
1230,
295,
661,
1880,
8077,
13,
759,
456,
307,
257,
636,
337,
50604,
50604,
1365,
11,
281,
483,
23551,
7144,
9214,
3587,
1412,
484,
295,
257,
13307,
2799,
11,
550,
321,
393,
360,
51052,
51052,
721,
411,
11,
1954,
11,
652,
988,
300,
264,
4195,
567,
307,
466,
281,
652,
257,
955,
14425,
1228,
341,
16599,
51284,
51284,
307,
300,
307,
411,
11,
1915,
2270,
10676,
264,
4195,
567,
7993,
2942,
341,
16599,
13,
407,
309,
575,
257,
1230,
51768,
51768
]
},
{
"avg_logprob": -0.1495452880859375,
"compression_ratio": 1.711111111111111,
"end": 504.48,
"id": 75,
"no_speech_prob": 0.00004129850276513025,
"seek": 49072,
"start": 495.52000000000004,
"temperature": 0,
"text": " example, to get reasonably secure authenticated data out of a smartphone camera, then we can do",
"tokens": [
50364,
4523,
484,
300,
341,
767,
9870,
257,
1230,
295,
661,
1880,
8077,
13,
759,
456,
307,
257,
636,
337,
50604,
50604,
1365,
11,
281,
483,
23551,
7144,
9214,
3587,
1412,
484,
295,
257,
13307,
2799,
11,
550,
321,
393,
360,
51052,
51052,
721,
411,
11,
1954,
11,
652,
988,
300,
264,
4195,
567,
307,
466,
281,
652,
257,
955,
14425,
1228,
341,
16599,
51284,
51284,
307,
300,
307,
411,
11,
1915,
2270,
10676,
264,
4195,
567,
7993,
2942,
341,
16599,
13,
407,
309,
575,
257,
1230,
51768,
51768
]
},
{
"avg_logprob": -0.1495452880859375,
"compression_ratio": 1.711111111111111,
"end": 509.12,
"id": 76,
"no_speech_prob": 0.00004129850276513025,
"seek": 49072,
"start": 504.48,
"temperature": 0,
"text": " things like, oh, make sure that the user who is about to make a big transaction using this wallet",
"tokens": [
50364,
4523,
484,
300,
341,
767,
9870,
257,
1230,
295,
661,
1880,
8077,
13,
759,
456,
307,
257,
636,
337,
50604,
50604,
1365,
11,
281,
483,
23551,
7144,
9214,
3587,
1412,
484,
295,
257,
13307,
2799,
11,
550,
321,
393,
360,
51052,
51052,
721,
411,
11,
1954,
11,
652,
988,
300,
264,
4195,
567,
307,
466,
281,
652,
257,
955,
14425,
1228,
341,
16599,
51284,
51284,
307,
300,
307,
411,
11,
1915,
2270,
10676,
264,
4195,
567,
7993,
2942,
341,
16599,
13,
407,
309,
575,
257,
1230,
51768,
51768
]
},
{
"avg_logprob": -0.1495452880859375,
"compression_ratio": 1.711111111111111,
"end": 518.8000000000001,
"id": 77,
"no_speech_prob": 0.00004129850276513025,
"seek": 49072,
"start": 509.12,
"temperature": 0,
"text": " is that is like, facially matches the user who originally created this wallet. So it has a number",
"tokens": [
50364,
4523,
484,
300,
341,
767,
9870,
257,
1230,
295,
661,
1880,
8077,
13,
759,
456,
307,
257,
636,
337,
50604,
50604,
1365,
11,
281,
483,
23551,
7144,
9214,
3587,
1412,
484,
295,
257,
13307,
2799,
11,
550,
321,
393,
360,
51052,
51052,
721,
411,
11,
1954,
11,
652,
988,
300,
264,
4195,
567,
307,
466,
281,
652,
257,
955,
14425,
1228,
341,
16599,
51284,
51284,
307,
300,
307,
411,
11,
1915,
2270,
10676,
264,
4195,
567,
7993,
2942,
341,
16599,
13,
407,
309,
575,
257,
1230,
51768,
51768
]
},
{
"avg_logprob": -0.1585106524554166,
"compression_ratio": 1.6724890829694323,
"end": 526.4799999999999,
"id": 78,
"no_speech_prob": 0.00023750841501168907,
"seek": 51880,
"start": 518.8,
"temperature": 0,
"text": " of like, security advantages and UX advantages we can achieve here. I think that's, and then there's",
"tokens": [
50364,
295,
411,
11,
3825,
14906,
293,
40176,
14906,
321,
393,
4584,
510,
13,
286,
519,
300,
311,
11,
293,
550,
456,
311,
50748,
50748,
411,
257,
3840,
295,
661,
721,
300,
321,
434,
12736,
689,
291,
393,
2995,
3228,
649,
10716,
2108,
819,
51024,
51024,
721,
411,
11,
291,
727,
2995,
15642,
3228,
649,
10716,
490,
264,
14715,
365,
15642,
3228,
649,
10716,
300,
366,
51380,
51380,
12187,
322,
2463,
4562,
8512,
11,
337,
1365,
13,
400,
309,
9870,
493,
257,
1230,
295,
1880,
51680,
51680
]
},
{
"avg_logprob": -0.1585106524554166,
"compression_ratio": 1.6724890829694323,
"end": 532,
"id": 79,
"no_speech_prob": 0.00023750841501168907,
"seek": 51880,
"start": 526.4799999999999,
"temperature": 0,
"text": " like a bunch of other things that we're exploring where you can match biometrics across different",
"tokens": [
50364,
295,
411,
11,
3825,
14906,
293,
40176,
14906,
321,
393,
4584,
510,
13,
286,
519,
300,
311,
11,
293,
550,
456,
311,
50748,
50748,
411,
257,
3840,
295,
661,
721,
300,
321,
434,
12736,
689,
291,
393,
2995,
3228,
649,
10716,
2108,
819,
51024,
51024,
721,
411,
11,
291,
727,
2995,
15642,
3228,
649,
10716,
490,
264,
14715,
365,
15642,
3228,
649,
10716,
300,
366,
51380,
51380,
12187,
322,
2463,
4562,
8512,
11,
337,
1365,
13,
400,
309,
9870,
493,
257,
1230,
295,
1880,
51680,
51680
]
},
{
"avg_logprob": -0.1585106524554166,
"compression_ratio": 1.6724890829694323,
"end": 539.12,
"id": 80,
"no_speech_prob": 0.00023750841501168907,
"seek": 51880,
"start": 532,
"temperature": 0,
"text": " things like, you could match facial biometrics from the orb with facial biometrics that are",
"tokens": [
50364,
295,
411,
11,
3825,
14906,
293,
40176,
14906,
321,
393,
4584,
510,
13,
286,
519,
300,
311,
11,
293,
550,
456,
311,
50748,
50748,
411,
257,
3840,
295,
661,
721,
300,
321,
434,
12736,
689,
291,
393,
2995,
3228,
649,
10716,
2108,
819,
51024,
51024,
721,
411,
11,
291,
727,
2995,
15642,
3228,
649,
10716,
490,
264,
14715,
365,
15642,
3228,
649,
10716,
300,
366,
51380,
51380,
12187,
322,
2463,
4562,
8512,
11,
337,
1365,
13,
400,
309,
9870,
493,
257,
1230,
295,
1880,
51680,
51680
]
},
{
"avg_logprob": -0.1585106524554166,
"compression_ratio": 1.6724890829694323,
"end": 545.12,
"id": 81,
"no_speech_prob": 0.00023750841501168907,
"seek": 51880,
"start": 539.12,
"temperature": 0,
"text": " stored on government digital documents, for example. And it opens up a number of interesting",
"tokens": [
50364,
295,
411,
11,
3825,
14906,
293,
40176,
14906,
321,
393,
4584,
510,
13,
286,
519,
300,
311,
11,
293,
550,
456,
311,
50748,
50748,
411,
257,
3840,
295,
661,
721,
300,
321,
434,
12736,
689,
291,
393,
2995,
3228,
649,
10716,
2108,
819,
51024,
51024,
721,
411,
11,
291,
727,
2995,
15642,
3228,
649,
10716,
490,
264,
14715,
365,
15642,
3228,
649,
10716,
300,
366,
51380,
51380,
12187,
322,
2463,
4562,
8512,
11,
337,
1365,
13,
400,
309,
9870,
493,
257,
1230,
295,
1880,
51680,
51680
]
},
{
"avg_logprob": -0.2960782766342163,
"compression_ratio": 1.7593582887700534,
"end": 551.12,
"id": 82,
"no_speech_prob": 0.000005681277343683178,
"seek": 54512,
"start": 545.12,
"temperature": 0,
"text": " use cases. Once you have access to authenticated sensor data, basically.",
"tokens": [
50364,
764,
3331,
13,
3443,
291,
362,
2105,
281,
9214,
3587,
10200,
1412,
11,
1936,
13,
50664,
50784,
1144,
291,
519,
300,
11,
286,
914,
11,
28235,
8837,
11,
28235,
300,
5759,
11,
291,
458,
11,
51136,
51220,
727,
312,
1238,
1152,
11,
291,
458,
11,
727,
312,
257,
13357,
13,
1144,
291,
519,
300,
16194,
300,
51420,
51460,
853,
281,
1884,
512,
1333,
295,
3832,
420,
1310,
512,
1269,
8837,
1715,
727,
312,
257,
13357,
30,
51860,
51860
]
},
{
"avg_logprob": -0.2960782766342163,
"compression_ratio": 1.7593582887700534,
"end": 560.5600000000001,
"id": 83,
"no_speech_prob": 0.000005681277343683178,
"seek": 54512,
"start": 553.52,
"temperature": 0,
"text": " Do you think that, I mean, trusting hardware, trusting that devices, you know,",
"tokens": [
50364,
764,
3331,
13,
3443,
291,
362,
2105,
281,
9214,
3587,
10200,
1412,
11,
1936,
13,
50664,
50784,
1144,
291,
519,
300,
11,
286,
914,
11,
28235,
8837,
11,
28235,
300,
5759,
11,
291,
458,
11,
51136,
51220,
727,
312,
1238,
1152,
11,
291,
458,
11,
727,
312,
257,
13357,
13,
1144,
291,
519,
300,
16194,
300,
51420,
51460,
853,
281,
1884,
512,
1333,
295,
3832,
420,
1310,
512,
1269,
8837,
1715,
727,
312,
257,
13357,
30,
51860,
51860
]
},
{
"avg_logprob": -0.2960782766342163,
"compression_ratio": 1.7593582887700534,
"end": 566.24,
"id": 84,
"no_speech_prob": 0.000005681277343683178,
"seek": 54512,
"start": 562.24,
"temperature": 0,
"text": " could be pretty hard, you know, could be a barrier. Do you think that initiatives that",
"tokens": [
50364,
764,
3331,
13,
3443,
291,
362,
2105,
281,
9214,
3587,
10200,
1412,
11,
1936,
13,
50664,
50784,
1144,
291,
519,
300,
11,
286,
914,
11,
28235,
8837,
11,
28235,
300,
5759,
11,
291,
458,
11,
51136,
51220,
727,
312,
1238,
1152,
11,
291,
458,
11,
727,
312,
257,
13357,
13,
1144,
291,
519,
300,
16194,
300,
51420,
51460,
853,
281,
1884,
512,
1333,
295,
3832,
420,
1310,
512,
1269,
8837,
1715,
727,
312,
257,
13357,
30,
51860,
51860
]
},
{
"avg_logprob": -0.2960782766342163,
"compression_ratio": 1.7593582887700534,
"end": 575.04,
"id": 85,
"no_speech_prob": 0.000005681277343683178,
"seek": 54512,
"start": 567.04,
"temperature": 0,
"text": " try to create some sort of standard or maybe some open hardware design could be a barrier?",
"tokens": [
50364,
764,
3331,
13,
3443,
291,
362,
2105,
281,
9214,
3587,
10200,
1412,
11,
1936,
13,
50664,
50784,
1144,
291,
519,
300,
11,
286,
914,
11,
28235,
8837,
11,
28235,
300,
5759,
11,
291,
458,
11,
51136,
51220,
727,
312,
1238,
1152,
11,
291,
458,
11,
727,
312,
257,
13357,
13,
1144,
291,
519,
300,
16194,
300,
51420,
51460,
853,
281,
1884,
512,
1333,
295,
3832,
420,
1310,
512,
1269,
8837,
1715,
727,
312,
257,
13357,
30,
51860,
51860
]
},
{
"avg_logprob": -0.1550079915941376,
"compression_ratio": 1.7285067873303168,
"end": 582.7199999999999,
"id": 86,
"no_speech_prob": 0.00006495795969385654,
"seek": 57504,
"start": 575.04,
"temperature": 0,
"text": " Could have to lower. So basically having people that, you know, build, you know, different",
"tokens": [
50364,
7497,
362,
281,
3126,
13,
407,
1936,
1419,
561,
300,
11,
291,
458,
11,
1322,
11,
291,
458,
11,
819,
50748,
50748,
3037,
295,
264,
912,
16034,
4302,
11,
293,
291,
458,
11,
8659,
437,
309,
775,
11,
420,
437,
727,
312,
264,
51040,
51040,
9482,
727,
5039,
420,
27336,
512,
295,
264,
11,
291,
458,
11,
3361,
17695,
300,
257,
4195,
11,
286,
914,
11,
51428,
51480,
815,
1851,
322,
11,
291,
458,
437,
11,
286,
362,
281,
3361,
13,
286,
914,
11,
321,
1217,
3361,
264,
2477,
47059,
562,
321,
11634,
51840,
51840
]
},
{
"avg_logprob": -0.1550079915941376,
"compression_ratio": 1.7285067873303168,
"end": 588.56,
"id": 87,
"no_speech_prob": 0.00006495795969385654,
"seek": 57504,
"start": 582.7199999999999,
"temperature": 0,
"text": " version of the same trusted device, and you know, effectively what it does, or what could be the",
"tokens": [
50364,
7497,
362,
281,
3126,
13,
407,
1936,
1419,
561,
300,
11,
291,
458,
11,
1322,
11,
291,
458,
11,
819,
50748,
50748,
3037,
295,
264,
912,
16034,
4302,
11,
293,
291,
458,
11,
8659,
437,
309,
775,
11,
420,
437,
727,
312,
264,
51040,
51040,
9482,
727,
5039,
420,
27336,
512,
295,
264,
11,
291,
458,
11,
3361,
17695,
300,
257,
4195,
11,
286,
914,
11,
51428,
51480,
815,
1851,
322,
11,
291,
458,
437,
11,
286,
362,
281,
3361,
13,
286,
914,
11,
321,
1217,
3361,
264,
2477,
47059,
562,
321,
11634,
51840,
51840
]
},
{
"avg_logprob": -0.1550079915941376,
"compression_ratio": 1.7285067873303168,
"end": 596.3199999999999,
"id": 88,
"no_speech_prob": 0.00006495795969385654,
"seek": 57504,
"start": 588.56,
"temperature": 0,
"text": " architecture could solve or mitigate some of the, you know, trust assumptions that a user, I mean,",
"tokens": [
50364,
7497,
362,
281,
3126,
13,
407,
1936,
1419,
561,
300,
11,
291,
458,
11,
1322,
11,
291,
458,
11,
819,
50748,
50748,
3037,
295,
264,
912,
16034,
4302,
11,
293,
291,
458,
11,
8659,
437,
309,
775,
11,
420,
437,
727,
312,
264,
51040,
51040,
9482,
727,
5039,
420,
27336,
512,
295,
264,
11,
291,
458,
11,
3361,
17695,
300,
257,
4195,
11,
286,
914,
11,
51428,
51480,
815,
1851,
322,
11,
291,
458,
437,
11,
286,
362,
281,
3361,
13,
286,
914,
11,
321,
1217,
3361,
264,
2477,
47059,
562,
321,
11634,
51840,
51840
]
},
{
"avg_logprob": -0.1550079915941376,
"compression_ratio": 1.7285067873303168,
"end": 604.56,
"id": 89,
"no_speech_prob": 0.00006495795969385654,
"seek": 57504,
"start": 597.36,
"temperature": 0,
"text": " may face on, you know what, I have to trust. I mean, we already trust the hyphen when we unlock",
"tokens": [
50364,
7497,
362,
281,
3126,
13,
407,
1936,
1419,
561,
300,
11,
291,
458,
11,
1322,
11,
291,
458,
11,
819,
50748,
50748,
3037,
295,
264,
912,
16034,
4302,
11,
293,
291,
458,
11,
8659,
437,
309,
775,
11,
420,
437,
727,
312,
264,
51040,
51040,
9482,
727,
5039,
420,
27336,
512,
295,
264,
11,
291,
458,
11,
3361,
17695,
300,
257,
4195,
11,
286,
914,
11,
51428,
51480,
815,
1851,
322,
11,
291,
458,
437,
11,
286,
362,
281,
3361,
13,
286,
914,
11,
321,
1217,
3361,
264,
2477,
47059,
562,
321,
11634,
51840,
51840
]
},
{
"avg_logprob": -0.11235557993253072,
"compression_ratio": 1.6743119266055047,
"end": 611.4399999999999,
"id": 90,
"no_speech_prob": 0.0001158377417596057,
"seek": 60456,
"start": 604.56,
"temperature": 0,
"text": " it. So it's just, you know, a matter of, you know, explaining that, but obviously,",
"tokens": [
50364,
309,
13,
407,
309,
311,
445,
11,
291,
458,
11,
257,
1871,
295,
11,
291,
458,
11,
13468,
300,
11,
457,
2745,
11,
50708,
50736,
1419,
512,
1269,
3832,
576,
11,
291,
458,
11,
854,
294,
341,
2020,
11,
412,
1935,
341,
307,
452,
1558,
13,
51040,
51040,
286,
914,
11,
321,
366,
5193,
264,
7787,
13,
440,
1154,
307,
406,
370,
709,
300,
309,
311,
1152,
337,
561,
281,
51456,
51456,
4445,
309,
13,
440,
1154,
307,
300,
291,
500,
380,
458,
300,
436,
767,
12270,
309,
8944,
13,
1743,
11,
51728,
51728
]
},
{
"avg_logprob": -0.11235557993253072,
"compression_ratio": 1.6743119266055047,
"end": 618.0799999999999,
"id": 91,
"no_speech_prob": 0.0001158377417596057,
"seek": 60456,
"start": 612,
"temperature": 0,
"text": " having some open standard would, you know, help in this sense, at least this is my idea.",
"tokens": [
50364,
309,
13,
407,
309,
311,
445,
11,
291,
458,
11,
257,
1871,
295,
11,
291,
458,
11,
13468,
300,
11,
457,
2745,
11,
50708,
50736,
1419,
512,
1269,
3832,
576,
11,
291,
458,
11,
854,
294,
341,
2020,
11,
412,
1935,
341,
307,
452,
1558,
13,
51040,
51040,
286,
914,
11,
321,
366,
5193,
264,
7787,
13,
440,
1154,
307,
406,
370,
709,
300,
309,
311,
1152,
337,
561,
281,
51456,
51456,
4445,
309,
13,
440,
1154,
307,
300,
291,
500,
380,
458,
300,
436,
767,
12270,
309,
8944,
13,
1743,
11,
51728,
51728
]
},
{
"avg_logprob": -0.11235557993253072,
"compression_ratio": 1.6743119266055047,
"end": 626.4,
"id": 92,
"no_speech_prob": 0.0001158377417596057,
"seek": 60456,
"start": 618.0799999999999,
"temperature": 0,
"text": " I mean, we are opening the standards. The problem is not so much that it's hard for people to",
"tokens": [
50364,
309,
13,
407,
309,
311,
445,
11,
291,
458,
11,
257,
1871,
295,
11,
291,
458,
11,
13468,
300,
11,
457,
2745,
11,
50708,
50736,
1419,
512,
1269,
3832,
576,
11,
291,
458,
11,
854,
294,
341,
2020,
11,
412,
1935,
341,
307,
452,
1558,
13,
51040,
51040,
286,
914,
11,
321,
366,
5193,
264,
7787,
13,
440,
1154,
307,
406,
370,
709,
300,
309,
311,
1152,
337,
561,
281,
51456,
51456,
4445,
309,
13,
440,
1154,
307,
300,
291,
500,
380,
458,
300,
436,
767,
12270,
309,
8944,
13,
1743,
11,
51728,
51728
]
},
{
"avg_logprob": -0.11235557993253072,
"compression_ratio": 1.6743119266055047,
"end": 631.8399999999999,
"id": 93,
"no_speech_prob": 0.0001158377417596057,
"seek": 60456,
"start": 626.4,
"temperature": 0,
"text": " implement it. The problem is that you don't know that they actually implemented it correctly. Like,",
"tokens": [
50364,
309,
13,
407,
309,
311,
445,
11,
291,
458,
11,
257,
1871,
295,
11,
291,
458,
11,
13468,
300,
11,
457,
2745,
11,
50708,
50736,
1419,
512,
1269,
3832,
576,
11,
291,
458,
11,
854,
294,
341,
2020,
11,
412,
1935,
341,
307,
452,
1558,
13,
51040,
51040,
286,
914,
11,
321,
366,
5193,
264,
7787,
13,
440,
1154,
307,
406,
370,
709,
300,
309,
311,
1152,
337,
561,
281,
51456,
51456,
4445,
309,
13,
440,
1154,
307,
300,
291,
500,
380,
458,
300,
436,
767,
12270,
309,
8944,
13,
1743,
11,
51728,
51728
]
},
{
"avg_logprob": -0.135371595621109,
"compression_ratio": 1.6592920353982301,
"end": 636.88,
"id": 94,
"no_speech_prob": 0.000014733809621247929,
"seek": 63184,
"start": 631.84,
"temperature": 0,
"text": " let's say you have an authenticating photo camera, and you create an open standard for that.",
"tokens": [
50364,
718,
311,
584,
291,
362,
364,
12466,
990,
5052,
2799,
11,
293,
291,
1884,
364,
1269,
3832,
337,
300,
13,
50616,
50676,
961,
311,
584,
257,
10316,
551,
689,
291,
445,
411,
11,
718,
311,
584,
11,
1465,
257,
361,
494,
70,
1228,
257,
462,
16508,
8886,
13397,
13,
51044,
51044,
823,
291,
528,
281,
1884,
512,
1333,
295,
1908,
25841,
295,
439,
264,
411,
2570,
3006,
8622,
13,
51344,
51344,
583,
577,
360,
291,
8539,
16055,
341,
25841,
30,
1743,
11,
577,
360,
291,
6909,
257,
1329,
295,
2570,
3006,
1908,
51592,
51592
]
},
{
"avg_logprob": -0.135371595621109,
"compression_ratio": 1.6592920353982301,
"end": 645.44,
"id": 95,
"no_speech_prob": 0.000014733809621247929,
"seek": 63184,
"start": 638.08,
"temperature": 0,
"text": " Let's say a dumb thing where you just like, let's say, sign a jpeg using a ECDSA signature.",
"tokens": [
50364,
718,
311,
584,
291,
362,
364,
12466,
990,
5052,
2799,
11,
293,
291,
1884,
364,
1269,
3832,
337,
300,
13,
50616,
50676,
961,
311,
584,
257,
10316,
551,
689,
291,
445,
411,
11,
718,
311,
584,
11,
1465,
257,
361,
494,
70,
1228,
257,
462,
16508,
8886,
13397,
13,
51044,
51044,
823,
291,
528,
281,
1884,
512,
1333,
295,
1908,
25841,
295,
439,
264,
411,
2570,
3006,
8622,
13,
51344,
51344,
583,
577,
360,
291,
8539,
16055,
341,
25841,
30,
1743,
11,
577,
360,
291,
6909,
257,
1329,
295,
2570,
3006,
1908,
51592,
51592
]
},
{
"avg_logprob": -0.135371595621109,
"compression_ratio": 1.6592920353982301,
"end": 651.44,
"id": 96,
"no_speech_prob": 0.000014733809621247929,
"seek": 63184,
"start": 645.44,
"temperature": 0,
"text": " Now you want to create some sort of public repository of all the like known correct cameras.",
"tokens": [
50364,
718,
311,
584,
291,
362,
364,
12466,
990,
5052,
2799,
11,
293,
291,
1884,
364,
1269,
3832,
337,
300,
13,
50616,
50676,
961,
311,
584,
257,
10316,
551,
689,
291,
445,
411,
11,
718,
311,
584,
11,
1465,
257,
361,
494,
70,
1228,
257,
462,
16508,
8886,
13397,
13,
51044,
51044,
823,
291,
528,
281,
1884,
512,
1333,
295,
1908,
25841,
295,
439,
264,
411,
2570,
3006,
8622,
13,
51344,
51344,
583,
577,
360,
291,
8539,
16055,
341,
25841,
30,
1743,
11,
577,
360,
291,
6909,
257,
1329,
295,
2570,
3006,
1908,
51592,
51592
]
},
{
"avg_logprob": -0.135371595621109,
"compression_ratio": 1.6592920353982301,
"end": 656.4000000000001,
"id": 97,
"no_speech_prob": 0.000014733809621247929,
"seek": 63184,
"start": 651.44,
"temperature": 0,
"text": " But how do you gatekeep this repository? Like, how do you maintain a list of known correct public",
"tokens": [
50364,
718,
311,
584,
291,
362,
364,
12466,
990,
5052,
2799,
11,
293,
291,
1884,
364,
1269,
3832,
337,
300,
13,
50616,
50676,
961,
311,
584,
257,
10316,
551,
689,
291,
445,
411,
11,
718,
311,
584,
11,
1465,
257,
361,
494,
70,
1228,
257,
462,
16508,
8886,
13397,
13,
51044,
51044,
823,
291,
528,
281,
1884,
512,
1333,
295,
1908,
25841,
295,
439,
264,
411,
2570,
3006,
8622,
13,
51344,
51344,
583,
577,
360,
291,
8539,
16055,
341,
25841,
30,
1743,
11,
577,
360,
291,
6909,
257,
1329,
295,
2570,
3006,
1908,
51592,
51592
]
},
{
"avg_logprob": -0.22885636302912346,
"compression_ratio": 1.6933333333333334,
"end": 663.76,
"id": 98,
"no_speech_prob": 0.00015094752598088235,
"seek": 65640,
"start": 656.4,
"temperature": 0,
"text": " keys? Yeah, I'll defend someone from... I mean, I understood. Yeah, you cannot attest that I did",
"tokens": [
50364,
9317,
30,
865,
11,
286,
603,
8602,
1580,
490,
485,
286,
914,
11,
286,
7320,
13,
865,
11,
291,
2644,
951,
377,
300,
286,
630,
50732,
50732,
264,
4001,
5759,
11,
291,
458,
11,
286,
914,
11,
1338,
11,
321,
366,
11,
291,
458,
11,
411,
264,
35619,
6873,
11,
291,
458,
11,
51024,
51024,
29753,
11,
1936,
11,
689,
264,
5772,
307,
45339,
294,
452,
3313,
11,
294,
452,
1575,
11,
2035,
13,
407,
300,
311,
264,
51352,
51352,
912,
1154,
300,
291,
362,
294,
661,
13,
583,
11,
291,
458,
11,
321,
362,
512,
1333,
295,
11,
291,
458,
11,
1217,
51804,
51804
]
},
{
"avg_logprob": -0.22885636302912346,
"compression_ratio": 1.6933333333333334,
"end": 669.6,
"id": 99,
"no_speech_prob": 0.00015094752598088235,
"seek": 65640,
"start": 663.76,
"temperature": 0,
"text": " the physical devices, you know, I mean, yeah, we are, you know, like the Eisenberg, you know,",
"tokens": [
50364,
9317,
30,
865,
11,
286,
603,
8602,
1580,
490,
485,
286,
914,
11,
286,
7320,
13,
865,
11,
291,
2644,
951,
377,
300,
286,
630,
50732,
50732,
264,
4001,
5759,
11,
291,
458,
11,
286,
914,
11,
1338,
11,
321,
366,
11,
291,
458,
11,
411,
264,
35619,
6873,
11,
291,
458,
11,
51024,
51024,
29753,
11,
1936,
11,
689,
264,
5772,
307,
45339,
294,
452,
3313,
11,
294,
452,
1575,
11,
2035,
13,
407,
300,
311,
264,
51352,
51352,
912,
1154,
300,
291,
362,
294,
661,
13,
583,
11,
291,
458,
11,
321,
362,
512,
1333,
295,
11,
291,
458,
11,
1217,
51804,
51804
]
},
{
"avg_logprob": -0.22885636302912346,
"compression_ratio": 1.6933333333333334,
"end": 676.16,
"id": 100,
"no_speech_prob": 0.00015094752598088235,
"seek": 65640,
"start": 669.6,
"temperature": 0,
"text": " microscope, basically, where the wave is collapsing in my eye, in my mind, whatever. So that's the",
"tokens": [
50364,
9317,
30,
865,
11,
286,
603,
8602,
1580,
490,
485,
286,
914,
11,
286,
7320,
13,
865,
11,
291,
2644,
951,
377,
300,
286,
630,
50732,
50732,
264,
4001,
5759,
11,
291,
458,
11,
286,
914,
11,
1338,
11,
321,
366,
11,
291,
458,
11,
411,
264,
35619,
6873,
11,
291,
458,
11,
51024,
51024,
29753,
11,
1936,
11,
689,
264,
5772,
307,
45339,
294,
452,
3313,
11,
294,
452,
1575,
11,
2035,
13,
407,
300,
311,
264,
51352,
51352,
912,
1154,
300,
291,
362,
294,
661,
13,
583,
11,
291,
458,
11,
321,
362,
512,
1333,
295,
11,
291,
458,
11,
1217,
51804,
51804
]
},
{
"avg_logprob": -0.22885636302912346,
"compression_ratio": 1.6933333333333334,
"end": 685.1999999999999,
"id": 101,
"no_speech_prob": 0.00015094752598088235,
"seek": 65640,
"start": 676.16,
"temperature": 0,
"text": " same problem that you have in other. But, you know, we have some sort of, you know, already",
"tokens": [
50364,
9317,
30,
865,
11,
286,
603,
8602,
1580,
490,
485,
286,
914,
11,
286,
7320,
13,
865,
11,
291,
2644,
951,
377,
300,
286,
630,
50732,
50732,
264,
4001,
5759,
11,
291,
458,
11,
286,
914,
11,
1338,
11,
321,
366,
11,
291,
458,
11,
411,
264,
35619,
6873,
11,
291,
458,
11,
51024,
51024,
29753,
11,
1936,
11,
689,
264,
5772,
307,
45339,
294,
452,
3313,
11,
294,
452,
1575,
11,
2035,
13,
407,
300,
311,
264,
51352,
51352,
912,
1154,
300,
291,
362,
294,
661,
13,
583,
11,
291,
458,
11,
321,
362,
512,
1333,
295,
11,
291,
458,
11,
1217,
51804,
51804
]
},
{
"avg_logprob": -0.19663215720135233,
"compression_ratio": 1.68,
"end": 692.08,
"id": 102,
"no_speech_prob": 0.00007029253902146593,
"seek": 68520,
"start": 685.2,
"temperature": 0,
"text": " some scheme about, you know, game theory for all of course, where we have a decentralized network of",
"tokens": [
50364,
512,
12232,
466,
11,
291,
458,
11,
1216,
5261,
337,
439,
295,
1164,
11,
689,
321,
362,
257,
32870,
3209,
295,
50708,
50756,
561,
11,
16667,
11,
12554,
11,
300,
1310,
362,
512,
3178,
294,
264,
1216,
281,
11,
291,
458,
11,
281,
360,
257,
1500,
51180,
51180,
300,
3494,
2277,
11,
291,
458,
11,
746,
13,
467,
1943,
380,
746,
300,
727,
5039,
13,
1610,
445,
5164,
51528,
51528,
5164,
3361,
5245,
11,
411,
415,
445,
848,
11,
321,
3361,
6373,
281,
360,
341,
1507,
8944,
562,
51760,
51760
]
},
{
"avg_logprob": -0.19663215720135233,
"compression_ratio": 1.68,
"end": 701.5200000000001,
"id": 103,
"no_speech_prob": 0.00007029253902146593,
"seek": 68520,
"start": 693.0400000000001,
"temperature": 0,
"text": " people, entities, agents, that maybe have some skin in the game to, you know, to do a test",
"tokens": [
50364,
512,
12232,
466,
11,
291,
458,
11,
1216,
5261,
337,
439,
295,
1164,
11,
689,
321,
362,
257,
32870,
3209,
295,
50708,
50756,
561,
11,
16667,
11,
12554,
11,
300,
1310,
362,
512,
3178,
294,
264,
1216,
281,
11,
291,
458,
11,
281,
360,
257,
1500,
51180,
51180,
300,
3494,
2277,
11,
291,
458,
11,
746,
13,
467,
1943,
380,
746,
300,
727,
5039,
13,
1610,
445,
5164,
51528,
51528,
5164,
3361,
5245,
11,
411,
415,
445,
848,
11,
321,
3361,
6373,
281,
360,
341,
1507,
8944,
562,
51760,
51760
]
},
{
"avg_logprob": -0.19663215720135233,
"compression_ratio": 1.68,
"end": 708.48,
"id": 104,
"no_speech_prob": 0.00007029253902146593,
"seek": 68520,
"start": 701.5200000000001,
"temperature": 0,
"text": " that truthfully, you know, something. It isn't something that could solve. Or just traditional",
"tokens": [
50364,
512,
12232,
466,
11,
291,
458,
11,
1216,
5261,
337,
439,
295,
1164,
11,
689,
321,
362,
257,
32870,
3209,
295,
50708,
50756,
561,
11,
16667,
11,
12554,
11,
300,
1310,
362,
512,
3178,
294,
264,
1216,
281,
11,
291,
458,
11,
281,
360,
257,
1500,
51180,
51180,
300,
3494,
2277,
11,
291,
458,
11,
746,
13,
467,
1943,
380,
746,
300,
727,
5039,
13,
1610,
445,
5164,
51528,
51528,
5164,
3361,
5245,
11,
411,
415,
445,
848,
11,
321,
3361,
6373,
281,
360,
341,
1507,
8944,
562,
51760,
51760
]
},
{
"avg_logprob": -0.19663215720135233,
"compression_ratio": 1.68,
"end": 713.12,
"id": 105,
"no_speech_prob": 0.00007029253902146593,
"seek": 68520,
"start": 708.48,
"temperature": 0,
"text": " traditional trust models, like he just said, we trust Apple to do this stuff correctly when",
"tokens": [
50364,
512,
12232,
466,
11,
291,
458,
11,
1216,
5261,
337,
439,
295,
1164,
11,
689,
321,
362,
257,
32870,
3209,
295,
50708,
50756,
561,
11,
16667,
11,
12554,
11,
300,
1310,
362,
512,
3178,
294,
264,
1216,
281,
11,
291,
458,
11,
281,
360,
257,
1500,
51180,
51180,
300,
3494,
2277,
11,
291,
458,
11,
746,
13,
467,
1943,
380,
746,
300,
727,
5039,
13,
1610,
445,
5164,
51528,
51528,
5164,
3361,
5245,
11,
411,
415,
445,
848,
11,
321,
3361,
6373,
281,
360,
341,
1507,
8944,
562,
51760,
51760
]
},
{
"avg_logprob": -0.12814978190830775,
"compression_ratio": 1.5684647302904564,
"end": 720.64,
"id": 106,
"no_speech_prob": 0.00007139764056773856,
"seek": 71312,
"start": 713.12,
"temperature": 0,
"text": " they do it. Like we trust, like if the iPhone through some hoops can produce authenticated",
"tokens": [
50364,
436,
360,
309,
13,
1743,
321,
3361,
11,
411,
498,
264,
7252,
807,
512,
1106,
3370,
393,
5258,
9214,
3587,
50740,
50740,
5267,
11,
550,
321,
576,
3361,
341,
281,
312,
3006,
13,
1436,
6373,
307,
886,
955,
281,
411,
2082,
493,
257,
4111,
51048,
51048,
411,
300,
13,
13157,
11,
286,
519,
558,
586,
11,
2139,
13969,
266,
420,
13575,
11,
286,
2870,
7101,
11,
457,
436,
611,
28964,
638,
51372,
51372,
294,
12466,
990,
8622,
13,
400,
337,
264,
3670,
18,
764,
1389,
11,
321,
3361,
552,
281,
360,
300,
8944,
11,
51640,
51640
]
},
{
"avg_logprob": -0.12814978190830775,
"compression_ratio": 1.5684647302904564,
"end": 726.8,
"id": 107,
"no_speech_prob": 0.00007139764056773856,
"seek": 71312,
"start": 720.64,
"temperature": 0,
"text": " images, then we would trust this to be correct. Because Apple is too big to like mess up a feature",
"tokens": [
50364,
436,
360,
309,
13,
1743,
321,
3361,
11,
411,
498,
264,
7252,
807,
512,
1106,
3370,
393,
5258,
9214,
3587,
50740,
50740,
5267,
11,
550,
321,
576,
3361,
341,
281,
312,
3006,
13,
1436,
6373,
307,
886,
955,
281,
411,
2082,
493,
257,
4111,
51048,
51048,
411,
300,
13,
13157,
11,
286,
519,
558,
586,
11,
2139,
13969,
266,
420,
13575,
11,
286,
2870,
7101,
11,
457,
436,
611,
28964,
638,
51372,
51372,
294,
12466,
990,
8622,
13,
400,
337,
264,
3670,
18,
764,
1389,
11,
321,
3361,
552,
281,
360,
300,
8944,
11,
51640,
51640
]
},
{
"avg_logprob": -0.12814978190830775,
"compression_ratio": 1.5684647302904564,
"end": 733.28,
"id": 108,
"no_speech_prob": 0.00007139764056773856,
"seek": 71312,
"start": 726.8,
"temperature": 0,
"text": " like that. Similarly, I think right now, either Nikon or Sony, I forget whom, but they also dabble",
"tokens": [
50364,
436,
360,
309,
13,
1743,
321,
3361,
11,
411,
498,
264,
7252,
807,
512,
1106,
3370,
393,
5258,
9214,
3587,
50740,
50740,
5267,
11,
550,
321,
576,
3361,
341,
281,
312,
3006,
13,
1436,
6373,
307,
886,
955,
281,
411,
2082,
493,
257,
4111,
51048,
51048,
411,
300,
13,
13157,
11,
286,
519,
558,
586,
11,
2139,
13969,
266,
420,
13575,
11,
286,
2870,
7101,
11,
457,
436,
611,
28964,
638,
51372,
51372,
294,
12466,
990,
8622,
13,
400,
337,
264,
3670,
18,
764,
1389,
11,
321,
3361,
552,
281,
360,
300,
8944,
11,
51640,
51640
]
},
{
"avg_logprob": -0.12814978190830775,
"compression_ratio": 1.5684647302904564,
"end": 738.64,
"id": 109,
"no_speech_prob": 0.00007139764056773856,
"seek": 71312,
"start": 733.28,
"temperature": 0,
"text": " in authenticating cameras. And for the web3 use case, we trust them to do that correctly,",
"tokens": [
50364,
436,
360,
309,
13,
1743,
321,
3361,
11,
411,
498,
264,
7252,
807,
512,
1106,
3370,
393,
5258,
9214,
3587,
50740,
50740,
5267,
11,
550,
321,
576,
3361,
341,
281,
312,
3006,
13,
1436,
6373,
307,
886,
955,
281,
411,
2082,
493,
257,
4111,
51048,
51048,
411,
300,
13,
13157,
11,
286,
519,
558,
586,
11,
2139,
13969,
266,
420,
13575,
11,
286,
2870,
7101,
11,
457,
436,
611,
28964,
638,
51372,
51372,
294,
12466,
990,
8622,
13,
400,
337,
264,
3670,
18,
764,
1389,
11,
321,
3361,
552,
281,
360,
300,
8944,
11,
51640,
51640
]
},
{
"avg_logprob": -0.0643328304948478,
"compression_ratio": 1.576271186440678,
"end": 744.56,
"id": 110,
"no_speech_prob": 0.00010225844744127244,
"seek": 73864,
"start": 738.64,
"temperature": 0,
"text": " because we basically trust people at Sony and Nikon to not even know what web3 is, let alone",
"tokens": [
50364,
570,
321,
1936,
3361,
561,
412,
13575,
293,
13969,
266,
281,
406,
754,
458,
437,
3670,
18,
307,
11,
718,
3312,
50660,
50660,
1322,
8837,
12382,
1208,
337,
641,
5121,
13,
1436,
558,
586,
11,
264,
2650,
307,
406,
456,
13,
50940,
50940,
11676,
307,
300,
382,
2321,
382,
264,
4786,
1813,
456,
11,
293,
721,
360,
483,
955,
1547,
11,
550,
51272,
51308,
561,
486,
767,
29085,
4404,
5847,
12626,
281,
2082,
926,
365,
341,
13,
639,
307,
11,
337,
1365,
11,
51620,
51620
]
},
{
"avg_logprob": -0.0643328304948478,
"compression_ratio": 1.576271186440678,
"end": 750.16,
"id": 111,
"no_speech_prob": 0.00010225844744127244,
"seek": 73864,
"start": 744.56,
"temperature": 0,
"text": " build hardware exploits for their benefit. Because right now, the opportunity is not there.",
"tokens": [
50364,
570,
321,
1936,
3361,
561,
412,
13575,
293,
13969,
266,
281,
406,
754,
458,
437,
3670,
18,
307,
11,
718,
3312,
50660,
50660,
1322,
8837,
12382,
1208,
337,
641,
5121,
13,
1436,
558,
586,
11,
264,
2650,
307,
406,
456,
13,
50940,
50940,
11676,
307,
300,
382,
2321,
382,
264,
4786,
1813,
456,
11,
293,
721,
360,
483,
955,
1547,
11,
550,
51272,
51308,
561,
486,
767,
29085,
4404,
5847,
12626,
281,
2082,
926,
365,
341,
13,
639,
307,
11,
337,
1365,
11,
51620,
51620
]
},
{
"avg_logprob": -0.0643328304948478,
"compression_ratio": 1.576271186440678,
"end": 756.8,
"id": 112,
"no_speech_prob": 0.00010225844744127244,
"seek": 73864,
"start": 750.16,
"temperature": 0,
"text": " Problem is that as soon as the opportunities become there, and things do get big enough, then",
"tokens": [
50364,
570,
321,
1936,
3361,
561,
412,
13575,
293,
13969,
266,
281,
406,
754,
458,
437,
3670,
18,
307,
11,
718,
3312,
50660,
50660,
1322,
8837,
12382,
1208,
337,
641,
5121,
13,
1436,
558,
586,
11,
264,
2650,
307,
406,
456,
13,
50940,
50940,
11676,
307,
300,
382,
2321,
382,
264,
4786,
1813,
456,
11,
293,
721,
360,
483,
955,
1547,
11,
550,
51272,
51308,
561,
486,
767,
29085,
4404,
5847,
12626,
281,
2082,
926,
365,
341,
13,
639,
307,
11,
337,
1365,
11,
51620,
51620
]
},
{
"avg_logprob": -0.0643328304948478,
"compression_ratio": 1.576271186440678,
"end": 763.76,
"id": 113,
"no_speech_prob": 0.00010225844744127244,
"seek": 73864,
"start": 757.52,
"temperature": 0,
"text": " people will actually infiltrate supply chains to mess around with this. This is, for example,",
"tokens": [
50364,
570,
321,
1936,
3361,
561,
412,
13575,
293,
13969,
266,
281,
406,
754,
458,
437,
3670,
18,
307,
11,
718,
3312,
50660,
50660,
1322,
8837,
12382,
1208,
337,
641,
5121,
13,
1436,
558,
586,
11,
264,
2650,
307,
406,
456,
13,
50940,
50940,
11676,
307,
300,
382,
2321,
382,
264,
4786,
1813,
456,
11,
293,
721,
360,
483,
955,
1547,
11,
550,
51272,
51308,
561,
486,
767,
29085,
4404,
5847,
12626,
281,
2082,
926,
365,
341,
13,
639,
307,
11,
337,
1365,
11,
51620,
51620
]
},
{
"avg_logprob": -0.09180345989408947,
"compression_ratio": 1.6228373702422145,
"end": 769.4399999999999,
"id": 114,
"no_speech_prob": 0.00007251373608596623,
"seek": 76376,
"start": 763.76,
"temperature": 0,
"text": " what is happening with decentralized exchanges, people literally apply for jobs at Verizon and",
"tokens": [
50364,
437,
307,
2737,
365,
32870,
27374,
11,
561,
3736,
3079,
337,
4782,
412,
44456,
293,
50648,
50648,
314,
12,
44,
13632,
11,
445,
370,
436,
483,
264,
33697,
281,
360,
24738,
1693,
2382,
445,
370,
436,
393,
747,
670,
9402,
50948,
50948,
412,
2416,
32395,
27374,
1228,
641,
732,
5952,
26643,
13,
10905,
2740,
13,
1042,
11,
51296,
51336,
264,
912,
551,
286,
576,
3811,
2314,
1564,
321,
352,
281,
16034,
14840,
11,
411,
9105,
11,
51600,
51600,
309,
486,
312,
439,
2055,
570,
264,
764,
1389,
307,
370,
1359,
300,
321,
500,
380,
362,
281,
3292,
466,
341,
13,
400,
51840,
51840
]
},
{
"avg_logprob": -0.09180345989408947,
"compression_ratio": 1.6228373702422145,
"end": 775.4399999999999,
"id": 115,
"no_speech_prob": 0.00007251373608596623,
"seek": 76376,
"start": 769.4399999999999,
"temperature": 0,
"text": " T-Mobile, just so they get the authorization to do SIM swaps just so they can take over accounts",
"tokens": [
50364,
437,
307,
2737,
365,
32870,
27374,
11,
561,
3736,
3079,
337,
4782,
412,
44456,
293,
50648,
50648,
314,
12,
44,
13632,
11,
445,
370,
436,
483,
264,
33697,
281,
360,
24738,
1693,
2382,
445,
370,
436,
393,
747,
670,
9402,
50948,
50948,
412,
2416,
32395,
27374,
1228,
641,
732,
5952,
26643,
13,
10905,
2740,
13,
1042,
11,
51296,
51336,
264,
912,
551,
286,
576,
3811,
2314,
1564,
321,
352,
281,
16034,
14840,
11,
411,
9105,
11,
51600,
51600,
309,
486,
312,
439,
2055,
570,
264,
764,
1389,
307,
370,
1359,
300,
321,
500,
380,
362,
281,
3292,
466,
341,
13,
400,
51840,
51840
]
},
{
"avg_logprob": -0.09180345989408947,
"compression_ratio": 1.6228373702422145,
"end": 782.4,
"id": 116,
"no_speech_prob": 0.00007251373608596623,
"seek": 76376,
"start": 775.4399999999999,
"temperature": 0,
"text": " at large centralized exchanges using their two factor authentication. Similar problems. Well,",
"tokens": [
50364,
437,
307,
2737,
365,
32870,
27374,
11,
561,
3736,
3079,
337,
4782,
412,
44456,
293,
50648,
50648,
314,
12,
44,
13632,
11,
445,
370,
436,
483,
264,
33697,
281,
360,
24738,
1693,
2382,
445,
370,
436,
393,
747,
670,
9402,
50948,
50948,
412,
2416,
32395,
27374,
1228,
641,
732,
5952,
26643,
13,
10905,
2740,
13,
1042,
11,
51296,
51336,
264,
912,
551,
286,
576,
3811,
2314,
1564,
321,
352,
281,
16034,
14840,
11,
411,
9105,
11,
51600,
51600,
309,
486,
312,
439,
2055,
570,
264,
764,
1389,
307,
370,
1359,
300,
321,
500,
380,
362,
281,
3292,
466,
341,
13,
400,
51840,
51840
]
},
{
"avg_logprob": -0.09180345989408947,
"compression_ratio": 1.6228373702422145,
"end": 788.48,
"id": 117,
"no_speech_prob": 0.00007251373608596623,
"seek": 76376,
"start": 783.2,
"temperature": 0,
"text": " the same thing I would imagine happens once we go to trusted sensors, like initially,",
"tokens": [
50364,
437,
307,
2737,
365,
32870,
27374,
11,
561,
3736,
3079,
337,
4782,
412,
44456,
293,
50648,
50648,
314,
12,
44,
13632,
11,
445,
370,
436,
483,
264,
33697,
281,
360,
24738,
1693,
2382,
445,
370,
436,
393,
747,
670,
9402,
50948,
50948,
412,
2416,
32395,
27374,
1228,
641,
732,
5952,
26643,
13,
10905,
2740,
13,
1042,
11,
51296,
51336,
264,
912,
551,
286,
576,
3811,
2314,
1564,
321,
352,
281,
16034,
14840,
11,
411,
9105,
11,
51600,
51600,
309,
486,
312,
439,
2055,
570,
264,
764,
1389,
307,
370,
1359,
300,
321,
500,
380,
362,
281,
3292,
466,
341,
13,
400,
51840,
51840
]
},
{
"avg_logprob": -0.09180345989408947,
"compression_ratio": 1.6228373702422145,
"end": 793.28,
"id": 118,
"no_speech_prob": 0.00007251373608596623,
"seek": 76376,
"start": 788.48,
"temperature": 0,
"text": " it will be all happy because the use case is so small that we don't have to worry about this. And",
"tokens": [
50364,
437,
307,
2737,
365,
32870,
27374,
11,
561,
3736,
3079,
337,
4782,
412,
44456,
293,
50648,
50648,
314,
12,
44,
13632,
11,
445,
370,
436,
483,
264,
33697,
281,
360,
24738,
1693,
2382,
445,
370,
436,
393,
747,
670,
9402,
50948,
50948,
412,
2416,
32395,
27374,
1228,
641,
732,
5952,
26643,
13,
10905,
2740,
13,
1042,
11,
51296,
51336,
264,
912,
551,
286,
576,
3811,
2314,
1564,
321,
352,
281,
16034,
14840,
11,
411,
9105,
11,
51600,
51600,
309,
486,
312,
439,
2055,
570,
264,
764,
1389,
307,
370,
1359,
300,
321,
500,
380,
362,
281,
3292,
466,
341,
13,
400,
51840,
51840
]
},
{
"avg_logprob": -0.17824210226535797,
"compression_ratio": 1.4863387978142077,
"end": 800.88,
"id": 119,
"no_speech_prob": 0.000018915379769168794,
"seek": 79328,
"start": 793.28,
"temperature": 0,
"text": " then the value, like assets on the management, so to say, will become large enough that it becomes",
"tokens": [
50364,
550,
264,
2158,
11,
411,
9769,
322,
264,
4592,
11,
370,
281,
584,
11,
486,
1813,
2416,
1547,
300,
309,
3643,
50744,
50744,
257,
22024,
5847,
5021,
2690,
13,
865,
11,
457,
321,
393,
3974,
264,
3148,
2108,
13,
759,
321,
366,
1075,
281,
51208,
51264,
1500,
34348,
294,
819,
2098,
1553,
40571,
322,
257,
2167,
935,
11,
1936,
11,
51704,
51704
]
},
{
"avg_logprob": -0.17824210226535797,
"compression_ratio": 1.4863387978142077,
"end": 810.16,
"id": 120,
"no_speech_prob": 0.000018915379769168794,
"seek": 79328,
"start": 800.88,
"temperature": 0,
"text": " a viable supply chain attack. Yeah, but we can spread the risk across. If we are able to",
"tokens": [
50364,
550,
264,
2158,
11,
411,
9769,
322,
264,
4592,
11,
370,
281,
584,
11,
486,
1813,
2416,
1547,
300,
309,
3643,
50744,
50744,
257,
22024,
5847,
5021,
2690,
13,
865,
11,
457,
321,
393,
3974,
264,
3148,
2108,
13,
759,
321,
366,
1075,
281,
51208,
51264,
1500,
34348,
294,
819,
2098,
1553,
40571,
322,
257,
2167,
935,
11,
1936,
11,
51704,
51704
]
},
{
"avg_logprob": -0.17824210226535797,
"compression_ratio": 1.4863387978142077,
"end": 820.0799999999999,
"id": 121,
"no_speech_prob": 0.000018915379769168794,
"seek": 79328,
"start": 811.28,
"temperature": 0,
"text": " test assessing in different ways without concentrating on a single point, basically,",
"tokens": [
50364,
550,
264,
2158,
11,
411,
9769,
322,
264,
4592,
11,
370,
281,
584,
11,
486,
1813,
2416,
1547,
300,
309,
3643,
50744,
50744,
257,
22024,
5847,
5021,
2690,
13,
865,
11,
457,
321,
393,
3974,
264,
3148,
2108,
13,
759,
321,
366,
1075,
281,
51208,
51264,
1500,
34348,
294,
819,
2098,
1553,
40571,
322,
257,
2167,
935,
11,
1936,
11,
51704,
51704
]
},
{
"avg_logprob": -0.1537288551899924,
"compression_ratio": 1.476923076923077,
"end": 829.6,
"id": 122,
"no_speech_prob": 0.00010503012890694663,
"seek": 82008,
"start": 820.08,
"temperature": 0,
"text": " the problem for SIM swap is that single accounts owes too much money that becomes very feasible to",
"tokens": [
50364,
264,
1154,
337,
24738,
18135,
307,
300,
2167,
9402,
50028,
886,
709,
1460,
300,
3643,
588,
26648,
281,
50840,
50840,
360,
300,
733,
295,
2690,
13,
583,
498,
321,
366,
1075,
281,
362,
11,
286,
500,
380,
1604,
567,
390,
1566,
11,
457,
51068,
51116,
3866,
2098,
11,
512,
1333,
295,
3866,
5952,
337,
26643,
300,
727,
312,
445,
294,
264,
2609,
11,
51592,
51680
]
},
{
"avg_logprob": -0.1537288551899924,
"compression_ratio": 1.476923076923077,
"end": 834.1600000000001,
"id": 123,
"no_speech_prob": 0.00010503012890694663,
"seek": 82008,
"start": 829.6,
"temperature": 0,
"text": " do that kind of attack. But if we are able to have, I don't remember who was saying, but",
"tokens": [
50364,
264,
1154,
337,
24738,
18135,
307,
300,
2167,
9402,
50028,
886,
709,
1460,
300,
3643,
588,
26648,
281,
50840,
50840,
360,
300,
733,
295,
2690,
13,
583,
498,
321,
366,
1075,
281,
362,
11,
286,
500,
380,
1604,
567,
390,
1566,
11,
457,
51068,
51116,
3866,
2098,
11,
512,
1333,
295,
3866,
5952,
337,
26643,
300,
727,
312,
445,
294,
264,
2609,
11,
51592,
51680
]
},
{
"avg_logprob": -0.1537288551899924,
"compression_ratio": 1.476923076923077,
"end": 844.6400000000001,
"id": 124,
"no_speech_prob": 0.00010503012890694663,
"seek": 82008,
"start": 835.12,
"temperature": 0,
"text": " multiple ways, some sort of multiple factor for authentication that could be just in the individual,",
"tokens": [
50364,
264,
1154,
337,
24738,
18135,
307,
300,
2167,
9402,
50028,
886,
709,
1460,
300,
3643,
588,
26648,
281,
50840,
50840,
360,
300,
733,
295,
2690,
13,
583,
498,
321,
366,
1075,
281,
362,
11,
286,
500,
380,
1604,
567,
390,
1566,
11,
457,
51068,
51116,
3866,
2098,
11,
512,
1333,
295,
3866,
5952,
337,
26643,
300,
727,
312,
445,
294,
264,
2609,
11,
51592,
51680
]
},
{
"avg_logprob": -0.15300264860454357,
"compression_ratio": 1.6710526315789473,
"end": 852.24,
"id": 125,
"no_speech_prob": 0.00022303276637103409,
"seek": 84464,
"start": 844.64,
"temperature": 0,
"text": " you know, web, an SMS or whatever, but also via schemes like social recovery or social",
"tokens": [
50364,
291,
458,
11,
3670,
11,
364,
38107,
420,
2035,
11,
457,
611,
5766,
26954,
411,
2093,
8597,
420,
2093,
50744,
50744,
951,
377,
399,
13,
286,
914,
11,
286,
458,
300,
36983,
6254,
575,
300,
1851,
11,
406,
570,
36983,
6254,
575,
445,
257,
2799,
11,
51044,
51044,
457,
570,
36983,
6254,
390,
294,
257,
3036,
365,
1071,
2146,
300,
286,
3361,
420,
294,
257,
1908,
2280,
420,
2035,
13,
51412,
51456,
407,
286,
519,
300,
1391,
498,
321,
362,
257,
8388,
689,
321,
393,
362,
439,
729,
3866,
7139,
295,
51856,
51856
]
},
{
"avg_logprob": -0.15300264860454357,
"compression_ratio": 1.6710526315789473,
"end": 858.24,
"id": 126,
"no_speech_prob": 0.00022303276637103409,
"seek": 84464,
"start": 852.24,
"temperature": 0,
"text": " attestation. I mean, I know that Emiliano has that face, not because Emiliano has just a camera,",
"tokens": [
50364,
291,
458,
11,
3670,
11,
364,
38107,
420,
2035,
11,
457,
611,
5766,
26954,
411,
2093,
8597,
420,
2093,
50744,
50744,
951,
377,
399,
13,
286,
914,
11,
286,
458,
300,
36983,
6254,
575,
300,
1851,
11,
406,
570,
36983,
6254,
575,
445,
257,
2799,
11,
51044,
51044,
457,
570,
36983,
6254,
390,
294,
257,
3036,
365,
1071,
2146,
300,
286,
3361,
420,
294,
257,
1908,
2280,
420,
2035,
13,
51412,
51456,
407,
286,
519,
300,
1391,
498,
321,
362,
257,
8388,
689,
321,
393,
362,
439,
729,
3866,
7139,
295,
51856,
51856
]
},
{
"avg_logprob": -0.15300264860454357,
"compression_ratio": 1.6710526315789473,
"end": 865.6,
"id": 127,
"no_speech_prob": 0.00022303276637103409,
"seek": 84464,
"start": 858.24,
"temperature": 0,
"text": " but because Emiliano was in a picture with another guy that I trust or in a public event or whatever.",
"tokens": [
50364,
291,
458,
11,
3670,
11,
364,
38107,
420,
2035,
11,
457,
611,
5766,
26954,
411,
2093,
8597,
420,
2093,
50744,
50744,
951,
377,
399,
13,
286,
914,
11,
286,
458,
300,
36983,
6254,
575,
300,
1851,
11,
406,
570,
36983,
6254,
575,
445,
257,
2799,
11,
51044,
51044,
457,
570,
36983,
6254,
390,
294,
257,
3036,
365,
1071,
2146,
300,
286,
3361,
420,
294,
257,
1908,
2280,
420,
2035,
13,
51412,
51456,
407,
286,
519,
300,
1391,
498,
321,
362,
257,
8388,
689,
321,
393,
362,
439,
729,
3866,
7139,
295,
51856,
51856
]
},
{
"avg_logprob": -0.15300264860454357,
"compression_ratio": 1.6710526315789473,
"end": 874.48,
"id": 128,
"no_speech_prob": 0.00022303276637103409,
"seek": 84464,
"start": 866.48,
"temperature": 0,
"text": " So I think that probably if we have a framework where we can have all those multiple sources of",
"tokens": [
50364,
291,
458,
11,
3670,
11,
364,
38107,
420,
2035,
11,
457,
611,
5766,
26954,
411,
2093,
8597,
420,
2093,
50744,
50744,
951,
377,
399,
13,
286,
914,
11,
286,
458,
300,
36983,
6254,
575,
300,
1851,
11,
406,
570,
36983,
6254,
575,
445,
257,
2799,
11,
51044,
51044,
457,
570,
36983,
6254,
390,
294,
257,
3036,
365,
1071,
2146,
300,
286,
3361,
420,
294,
257,
1908,
2280,
420,
2035,
13,
51412,
51456,
407,
286,
519,
300,
1391,
498,
321,
362,
257,
8388,
689,
321,
393,
362,
439,
729,
3866,
7139,
295,
51856,
51856
]
},
{
"avg_logprob": -0.21624280535985554,
"compression_ratio": 1.5082872928176796,
"end": 884.48,
"id": 129,
"no_speech_prob": 0.0008467821171507239,
"seek": 87448,
"start": 874.48,
"temperature": 0,
"text": " data with the respective way for saying that something is true, it could work somehow,",
"tokens": [
50364,
1412,
365,
264,
23649,
636,
337,
1566,
300,
746,
307,
2074,
11,
309,
727,
589,
6063,
11,
50864,
50904,
2831,
813,
2121,
3319,
294,
2685,
2899,
13,
639,
307,
562,
291,
483,
281,
264,
917,
1216,
295,
1936,
51336,
51336,
445,
884,
1577,
322,
538,
1858,
293,
3148,
5215,
295,
411,
11,
4177,
11,
341,
954,
307,
19232,
341,
51584,
51584
]
},
{
"avg_logprob": -0.21624280535985554,
"compression_ratio": 1.5082872928176796,
"end": 893.9200000000001,
"id": 130,
"no_speech_prob": 0.0008467821171507239,
"seek": 87448,
"start": 885.28,
"temperature": 0,
"text": " rather than specializing in specific technology. This is when you get to the end game of basically",
"tokens": [
50364,
1412,
365,
264,
23649,
636,
337,
1566,
300,
746,
307,
2074,
11,
309,
727,
589,
6063,
11,
50864,
50904,
2831,
813,
2121,
3319,
294,
2685,
2899,
13,
639,
307,
562,
291,
483,
281,
264,
917,
1216,
295,
1936,
51336,
51336,
445,
884,
1577,
322,
538,
1858,
293,
3148,
5215,
295,
411,
11,
4177,
11,
341,
954,
307,
19232,
341,
51584,
51584
]
},
{
"avg_logprob": -0.21624280535985554,
"compression_ratio": 1.5082872928176796,
"end": 898.88,
"id": 131,
"no_speech_prob": 0.0008467821171507239,
"seek": 87448,
"start": 893.9200000000001,
"temperature": 0,
"text": " just doing full on by easy and risk analysis of like, hey, this person is claiming this",
"tokens": [
50364,
1412,
365,
264,
23649,
636,
337,
1566,
300,
746,
307,
2074,
11,
309,
727,
589,
6063,
11,
50864,
50904,
2831,
813,
2121,
3319,
294,
2685,
2899,
13,
639,
307,
562,
291,
483,
281,
264,
917,
1216,
295,
1936,
51336,
51336,
445,
884,
1577,
322,
538,
1858,
293,
3148,
5215,
295,
411,
11,
4177,
11,
341,
954,
307,
19232,
341,
51584,
51584
]
},
{
"avg_logprob": -0.09963176033713601,
"compression_ratio": 1.6354166666666667,
"end": 906.48,
"id": 132,
"no_speech_prob": 0.0002950624329969287,
"seek": 89888,
"start": 898.88,
"temperature": 0,
"text": " authorization with this evidence, like, what is the probability that this is incorrect? And then",
"tokens": [
50364,
33697,
365,
341,
4467,
11,
411,
11,
437,
307,
264,
8482,
300,
341,
307,
18424,
30,
400,
550,
50744,
50744,
291,
584,
11,
731,
11,
337,
1359,
16856,
11,
286,
478,
4950,
281,
25773,
411,
257,
502,
4,
7908,
3353,
3314,
13,
400,
51072,
51112,
286,
478,
754,
4950,
281,
411,
829,
452,
1460,
322,
264,
1622,
293,
1419,
281,
41685,
341,
954,
13,
51336,
51336,
13813,
337,
588,
2416,
16856,
11,
291,
528,
1310,
2946,
1256,
374,
2676,
11,
293,
291,
393,
19719,
51564,
51564,
5176,
2158,
293,
12288,
11,
12288,
13,
639,
307,
1936,
437,
955,
10237,
366,
4362,
884,
11,
597,
307,
983,
51848,
51848
]
},
{
"avg_logprob": -0.09963176033713601,
"compression_ratio": 1.6354166666666667,
"end": 913.04,
"id": 133,
"no_speech_prob": 0.0002950624329969287,
"seek": 89888,
"start": 906.48,
"temperature": 0,
"text": " you say, well, for small transactions, I'm willing to tolerate like a 1% false positive rate. And",
"tokens": [
50364,
33697,
365,
341,
4467,
11,
411,
11,
437,
307,
264,
8482,
300,
341,
307,
18424,
30,
400,
550,
50744,
50744,
291,
584,
11,
731,
11,
337,
1359,
16856,
11,
286,
478,
4950,
281,
25773,
411,
257,
502,
4,
7908,
3353,
3314,
13,
400,
51072,
51112,
286,
478,
754,
4950,
281,
411,
829,
452,
1460,
322,
264,
1622,
293,
1419,
281,
41685,
341,
954,
13,
51336,
51336,
13813,
337,
588,
2416,
16856,
11,
291,
528,
1310,
2946,
1256,
374,
2676,
11,
293,
291,
393,
19719,
51564,
51564,
5176,
2158,
293,
12288,
11,
12288,
13,
639,
307,
1936,
437,
955,
10237,
366,
4362,
884,
11,
597,
307,
983,
51848,
51848
]
},
{
"avg_logprob": -0.09963176033713601,
"compression_ratio": 1.6354166666666667,
"end": 918.32,
"id": 134,
"no_speech_prob": 0.0002950624329969287,
"seek": 89888,
"start": 913.84,
"temperature": 0,
"text": " I'm even willing to like put my money on the line and having to reimburse this person.",
"tokens": [
50364,
33697,
365,
341,
4467,
11,
411,
11,
437,
307,
264,
8482,
300,
341,
307,
18424,
30,
400,
550,
50744,
50744,
291,
584,
11,
731,
11,
337,
1359,
16856,
11,
286,
478,
4950,
281,
25773,
411,
257,
502,
4,
7908,
3353,
3314,
13,
400,
51072,
51112,
286,
478,
754,
4950,
281,
411,
829,
452,
1460,
322,
264,
1622,
293,
1419,
281,
41685,
341,
954,
13,
51336,
51336,
13813,
337,
588,
2416,
16856,
11,
291,
528,
1310,
2946,
1256,
374,
2676,
11,
293,
291,
393,
19719,
51564,
51564,
5176,
2158,
293,
12288,
11,
12288,
13,
639,
307,
1936,
437,
955,
10237,
366,
4362,
884,
11,
597,
307,
983,
51848,
51848
]
},
{
"avg_logprob": -0.09963176033713601,
"compression_ratio": 1.6354166666666667,
"end": 922.88,
"id": 135,
"no_speech_prob": 0.0002950624329969287,
"seek": 89888,
"start": 918.32,
"temperature": 0,
"text": " Whereas for very large transactions, you want maybe higher assurances, and you can optimize",
"tokens": [
50364,
33697,
365,
341,
4467,
11,
411,
11,
437,
307,
264,
8482,
300,
341,
307,
18424,
30,
400,
550,
50744,
50744,
291,
584,
11,
731,
11,
337,
1359,
16856,
11,
286,
478,
4950,
281,
25773,
411,
257,
502,
4,
7908,
3353,
3314,
13,
400,
51072,
51112,
286,
478,
754,
4950,
281,
411,
829,
452,
1460,
322,
264,
1622,
293,
1419,
281,
41685,
341,
954,
13,
51336,
51336,
13813,
337,
588,
2416,
16856,
11,
291,
528,
1310,
2946,
1256,
374,
2676,
11,
293,
291,
393,
19719,
51564,
51564,
5176,
2158,
293,
12288,
11,
12288,
13,
639,
307,
1936,
437,
955,
10237,
366,
4362,
884,
11,
597,
307,
983,
51848,
51848
]
},
{
"avg_logprob": -0.09963176033713601,
"compression_ratio": 1.6354166666666667,
"end": 928.56,
"id": 136,
"no_speech_prob": 0.0002950624329969287,
"seek": 89888,
"start": 922.88,
"temperature": 0,
"text": " expected value and blah, blah. This is basically what big banks are currently doing, which is why",
"tokens": [
50364,
33697,
365,
341,
4467,
11,
411,
11,
437,
307,
264,
8482,
300,
341,
307,
18424,
30,
400,
550,
50744,
50744,
291,
584,
11,
731,
11,
337,
1359,
16856,
11,
286,
478,
4950,
281,
25773,
411,
257,
502,
4,
7908,
3353,
3314,
13,
400,
51072,
51112,
286,
478,
754,
4950,
281,
411,
829,
452,
1460,
322,
264,
1622,
293,
1419,
281,
41685,
341,
954,
13,
51336,
51336,
13813,
337,
588,
2416,
16856,
11,
291,
528,
1310,
2946,
1256,
374,
2676,
11,
293,
291,
393,
19719,
51564,
51564,
5176,
2158,
293,
12288,
11,
12288,
13,
639,
307,
1936,
437,
955,
10237,
366,
4362,
884,
11,
597,
307,
983,
51848,
51848
]
},
{
"avg_logprob": -0.11475808741682667,
"compression_ratio": 1.6095890410958904,
"end": 933.3599999999999,
"id": 137,
"no_speech_prob": 0.000049081354518420994,
"seek": 92856,
"start": 928.56,
"temperature": 0,
"text": " you can get away with so many crappy things that the banks are doing, because they have these risk",
"tokens": [
50364,
291,
393,
483,
1314,
365,
370,
867,
36531,
721,
300,
264,
10237,
366,
884,
11,
570,
436,
362,
613,
3148,
50604,
50604,
5245,
293,
436,
445,
458,
437,
264,
4923,
39457,
366,
13,
10635,
551,
365,
3329,
13,
400,
309,
311,
411,
33636,
18238,
4983,
11,
50948,
50948,
689,
411,
2171,
309,
311,
257,
2199,
1520,
4995,
11,
2171,
291,
643,
281,
352,
807,
544,
1106,
3370,
13,
51176,
51176,
467,
311,
1128,
534,
1850,
983,
293,
577,
11,
457,
309,
311,
445,
2361,
322,
641,
646,
917,
3148,
5245,
13,
51380,
51420,
440,
912,
551,
307,
2074,
337,
7214,
382,
731,
11,
558,
30,
1743,
561,
362,
257,
2946,
12578,
295,
8177,
13,
51668,
51668
]
},
{
"avg_logprob": -0.11475808741682667,
"compression_ratio": 1.6095890410958904,
"end": 940.2399999999999,
"id": 138,
"no_speech_prob": 0.000049081354518420994,
"seek": 92856,
"start": 933.3599999999999,
"temperature": 0,
"text": " models and they just know what the trade offs are. Same thing with Google. And it's like CAPTCHA,",
"tokens": [
50364,
291,
393,
483,
1314,
365,
370,
867,
36531,
721,
300,
264,
10237,
366,
884,
11,
570,
436,
362,
613,
3148,
50604,
50604,
5245,
293,
436,
445,
458,
437,
264,
4923,
39457,
366,
13,
10635,
551,
365,
3329,
13,
400,
309,
311,
411,
33636,
18238,
4983,
11,
50948,
50948,
689,
411,
2171,
309,
311,
257,
2199,
1520,
4995,
11,
2171,
291,
643,
281,
352,
807,
544,
1106,
3370,
13,
51176,
51176,
467,
311,
1128,
534,
1850,
983,
293,
577,
11,
457,
309,
311,
445,
2361,
322,
641,
646,
917,
3148,
5245,
13,
51380,
51420,
440,
912,
551,
307,
2074,
337,
7214,
382,
731,
11,
558,
30,
1743,
561,
362,
257,
2946,
12578,
295,
8177,
13,
51668,
51668
]
},
{
"avg_logprob": -0.11475808741682667,
"compression_ratio": 1.6095890410958904,
"end": 944.8,
"id": 139,
"no_speech_prob": 0.000049081354518420994,
"seek": 92856,
"start": 940.2399999999999,
"temperature": 0,
"text": " where like sometimes it's a simple checkbox, sometimes you need to go through more hoops.",
"tokens": [
50364,
291,
393,
483,
1314,
365,
370,
867,
36531,
721,
300,
264,
10237,
366,
884,
11,
570,
436,
362,
613,
3148,
50604,
50604,
5245,
293,
436,
445,
458,
437,
264,
4923,
39457,
366,
13,
10635,
551,
365,
3329,
13,
400,
309,
311,
411,
33636,
18238,
4983,
11,
50948,
50948,
689,
411,
2171,
309,
311,
257,
2199,
1520,
4995,
11,
2171,
291,
643,
281,
352,
807,
544,
1106,
3370,
13,
51176,
51176,
467,
311,
1128,
534,
1850,
983,
293,
577,
11,
457,
309,
311,
445,
2361,
322,
641,
646,
917,
3148,
5245,
13,
51380,
51420,
440,
912,
551,
307,
2074,
337,
7214,
382,
731,
11,
558,
30,
1743,
561,
362,
257,
2946,
12578,
295,
8177,
13,
51668,
51668
]
},
{
"avg_logprob": -0.11475808741682667,
"compression_ratio": 1.6095890410958904,
"end": 948.88,
"id": 140,
"no_speech_prob": 0.000049081354518420994,
"seek": 92856,
"start": 944.8,
"temperature": 0,
"text": " It's never really clear why and how, but it's just based on their back end risk models.",
"tokens": [
50364,
291,
393,
483,
1314,
365,
370,
867,
36531,
721,
300,
264,
10237,
366,
884,
11,
570,
436,
362,
613,
3148,
50604,
50604,
5245,
293,
436,
445,
458,
437,
264,
4923,
39457,
366,
13,
10635,
551,
365,
3329,
13,
400,
309,
311,
411,
33636,
18238,
4983,
11,
50948,
50948,
689,
411,
2171,
309,
311,
257,
2199,
1520,
4995,
11,
2171,
291,
643,
281,
352,
807,
544,
1106,
3370,
13,
51176,
51176,
467,
311,
1128,
534,
1850,
983,
293,
577,
11,
457,
309,
311,
445,
2361,
322,
641,
646,
917,
3148,
5245,
13,
51380,
51420,
440,
912,
551,
307,
2074,
337,
7214,
382,
731,
11,
558,
30,
1743,
561,
362,
257,
2946,
12578,
295,
8177,
13,
51668,
51668
]
},
{
"avg_logprob": -0.11475808741682667,
"compression_ratio": 1.6095890410958904,
"end": 954.64,
"id": 141,
"no_speech_prob": 0.000049081354518420994,
"seek": 92856,
"start": 949.68,
"temperature": 0,
"text": " The same thing is true for insurance as well, right? Like people have a higher burden of proof.",
"tokens": [
50364,
291,
393,
483,
1314,
365,
370,
867,
36531,
721,
300,
264,
10237,
366,
884,
11,
570,
436,
362,
613,
3148,
50604,
50604,
5245,
293,
436,
445,
458,
437,
264,
4923,
39457,
366,
13,
10635,
551,
365,
3329,
13,
400,
309,
311,
411,
33636,
18238,
4983,
11,
50948,
50948,
689,
411,
2171,
309,
311,
257,
2199,
1520,
4995,
11,
2171,
291,
643,
281,
352,
807,
544,
1106,
3370,
13,
51176,
51176,
467,
311,
1128,
534,
1850,
983,
293,
577,
11,
457,
309,
311,
445,
2361,
322,
641,
646,
917,
3148,
5245,
13,
51380,
51420,
440,
912,
551,
307,
2074,
337,
7214,
382,
731,
11,
558,
30,
1743,
561,
362,
257,
2946,
12578,
295,
8177,
13,
51668,
51668
]
},
{
"avg_logprob": -0.1275254816248797,
"compression_ratio": 1.5561797752808988,
"end": 960.24,
"id": 142,
"no_speech_prob": 0.000012409998817020096,
"seek": 95464,
"start": 954.64,
"temperature": 0,
"text": " It's a more substantial claim, which is also like, we talk about end game, like some of this,",
"tokens": [
50364,
467,
311,
257,
544,
16726,
3932,
11,
597,
307,
611,
411,
11,
321,
751,
466,
917,
1216,
11,
411,
512,
295,
341,
11,
50644,
50704,
16599,
8597,
1507,
11,
382,
731,
382,
1306,
5489,
7214,
9441,
307,
8344,
411,
264,
917,
1216,
51080,
51080,
337,
512,
295,
613,
5245,
382,
731,
13,
286,
914,
11,
445,
5994,
11,
5994,
1606,
300,
393,
380,
2514,
322,
51344,
51344
]
},
{
"avg_logprob": -0.1275254816248797,
"compression_ratio": 1.5561797752808988,
"end": 968.96,
"id": 143,
"no_speech_prob": 0.000012409998817020096,
"seek": 95464,
"start": 961.4399999999999,
"temperature": 0,
"text": " wallet recovery stuff, as well as verifying insurance claims is somewhat like the end game",
"tokens": [
50364,
467,
311,
257,
544,
16726,
3932,
11,
597,
307,
611,
411,
11,
321,
751,
466,
917,
1216,
11,
411,
512,
295,
341,
11,
50644,
50704,
16599,
8597,
1507,
11,
382,
731,
382,
1306,
5489,
7214,
9441,
307,
8344,
411,
264,
917,
1216,
51080,
51080,
337,
512,
295,
613,
5245,
382,
731,
13,
286,
914,
11,
445,
5994,
11,
5994,
1606,
300,
393,
380,
2514,
322,
51344,
51344
]
},
{
"avg_logprob": -0.1275254816248797,
"compression_ratio": 1.5561797752808988,
"end": 974.24,
"id": 144,
"no_speech_prob": 0.000012409998817020096,
"seek": 95464,
"start": 968.96,
"temperature": 0,
"text": " for some of these models as well. I mean, just massive, massive business that can't exist on",
"tokens": [
50364,
467,
311,
257,
544,
16726,
3932,
11,
597,
307,
611,
411,
11,
321,
751,
466,
917,
1216,
11,
411,
512,
295,
341,
11,
50644,
50704,
16599,
8597,
1507,
11,
382,
731,
382,
1306,
5489,
7214,
9441,
307,
8344,
411,
264,
917,
1216,
51080,
51080,
337,
512,
295,
613,
5245,
382,
731,
13,
286,
914,
11,
445,
5994,
11,
5994,
1606,
300,
393,
380,
2514,
322,
51344,
51344
]
},
{
"avg_logprob": -0.15157823395310788,
"compression_ratio": 1.4037267080745341,
"end": 985.76,
"id": 145,
"no_speech_prob": 0.000055601125495741144,
"seek": 97424,
"start": 974.24,
"temperature": 0,
"text": " chain because there's no good verification of claims, but ML could be one tool. Cool.",
"tokens": [
50364,
5021,
570,
456,
311,
572,
665,
30206,
295,
9441,
11,
457,
21601,
727,
312,
472,
2290,
13,
8561,
13,
50940,
51044,
492,
445,
362,
445,
337,
264,
9717,
295,
565,
11,
1732,
2077,
13,
286,
519,
300,
321,
393,
2897,
670,
51376,
51428,
472,
4829,
1936,
420,
512,
1524,
493,
1168,
281,
341,
13,
51732
]
},
{
"avg_logprob": -0.15157823395310788,
"compression_ratio": 1.4037267080745341,
"end": 994.48,
"id": 146,
"no_speech_prob": 0.000055601125495741144,
"seek": 97424,
"start": 987.84,
"temperature": 0,
"text": " We just have just for the sake of time, five minutes. I think that we can stick over",
"tokens": [
50364,
5021,
570,
456,
311,
572,
665,
30206,
295,
9441,
11,
457,
21601,
727,
312,
472,
2290,
13,
8561,
13,
50940,
51044,
492,
445,
362,
445,
337,
264,
9717,
295,
565,
11,
1732,
2077,
13,
286,
519,
300,
321,
393,
2897,
670,
51376,
51428,
472,
4829,
1936,
420,
512,
1524,
493,
1168,
281,
341,
13,
51732
]
},
{
"avg_logprob": -0.2548191170943411,
"compression_ratio": 1.4385964912280702,
"end": 1003.52,
"id": 147,
"no_speech_prob": 0.00012143708590883762,
"seek": 99448,
"start": 994.48,
"temperature": 0,
"text": " one topic basically or some follow up question to this.",
"tokens": [
50364,
472,
4829,
1936,
420,
512,
1524,
493,
1168,
281,
341,
13,
50816,
50916,
865,
11,
1310,
257,
1168,
281,
264,
1594,
510,
13,
4402,
2878,
362,
257,
665,
12492,
7222,
420,
445,
257,
51284,
51284,
14089,
754,
295,
439,
264,
819,
3467,
295,
2098,
321,
393,
360,
36663,
3479,
2539,
30,
407,
1953,
51728,
51728
]
},
{
"avg_logprob": -0.2548191170943411,
"compression_ratio": 1.4385964912280702,
"end": 1012.88,
"id": 148,
"no_speech_prob": 0.00012143708590883762,
"seek": 99448,
"start": 1005.52,
"temperature": 0,
"text": " Yeah, maybe a question to the group here. Does anyone have a good overview article or just a",
"tokens": [
50364,
472,
4829,
1936,
420,
512,
1524,
493,
1168,
281,
341,
13,
50816,
50916,
865,
11,
1310,
257,
1168,
281,
264,
1594,
510,
13,
4402,
2878,
362,
257,
665,
12492,
7222,
420,
445,
257,
51284,
51284,
14089,
754,
295,
439,
264,
819,
3467,
295,
2098,
321,
393,
360,
36663,
3479,
2539,
30,
407,
1953,
51728,
51728
]
},
{
"avg_logprob": -0.2548191170943411,
"compression_ratio": 1.4385964912280702,
"end": 1021.76,
"id": 149,
"no_speech_prob": 0.00012143708590883762,
"seek": 99448,
"start": 1012.88,
"temperature": 0,
"text": " graphic even of all the different types of ways we can do encrypted machine learning? So thinking",
"tokens": [
50364,
472,
4829,
1936,
420,
512,
1524,
493,
1168,
281,
341,
13,
50816,
50916,
865,
11,
1310,
257,
1168,
281,
264,
1594,
510,
13,
4402,
2878,
362,
257,
665,
12492,
7222,
420,
445,
257,
51284,
51284,
14089,
754,
295,
439,
264,
819,
3467,
295,
2098,
321,
393,
360,
36663,
3479,
2539,
30,
407,
1953,
51728,
51728
]
},
{
"avg_logprob": -0.37404739972457146,
"compression_ratio": 1.5743801652892562,
"end": 1030,
"id": 150,
"no_speech_prob": 0.00012728571891784668,
"seek": 102176,
"start": 1021.76,
"temperature": 0,
"text": " about a mapping of FHE, of federated learning, differential privacy, and including ZKML and all",
"tokens": [
50364,
466,
257,
18350,
295,
479,
23728,
11,
295,
38024,
770,
2539,
11,
15756,
11427,
11,
293,
3009,
1176,
42,
12683,
293,
439,
50776,
50776,
1080,
2098,
293,
6422,
13,
1119,
456,
257,
665,
439,
28268,
278,
5934,
337,
341,
4992,
30,
51104,
51244,
1610,
820,
321,
829,
309,
1214,
30,
286,
500,
380,
519,
456,
311,
364,
16961,
295,
439,
729,
7909,
13,
286,
458,
300,
456,
311,
51476,
51476,
512,
479,
23728,
1507,
13,
286,
600,
829,
1214,
257,
5934,
1219,
611,
1176,
42,
12683,
11,
597,
307,
411,
11,
286,
727,
1884,
1176,
42,
12683,
1507,
13,
51804,
51804
]
},
{
"avg_logprob": -0.37404739972457146,
"compression_ratio": 1.5743801652892562,
"end": 1036.56,
"id": 151,
"no_speech_prob": 0.00012728571891784668,
"seek": 102176,
"start": 1030,
"temperature": 0,
"text": " its ways and forms. Is there a good all encompassing guide for this anywhere?",
"tokens": [
50364,
466,
257,
18350,
295,
479,
23728,
11,
295,
38024,
770,
2539,
11,
15756,
11427,
11,
293,
3009,
1176,
42,
12683,
293,
439,
50776,
50776,
1080,
2098,
293,
6422,
13,
1119,
456,
257,
665,
439,
28268,
278,
5934,
337,
341,
4992,
30,
51104,
51244,
1610,
820,
321,
829,
309,
1214,
30,
286,
500,
380,
519,
456,
311,
364,
16961,
295,
439,
729,
7909,
13,
286,
458,
300,
456,
311,
51476,
51476,
512,
479,
23728,
1507,
13,
286,
600,
829,
1214,
257,
5934,
1219,
611,
1176,
42,
12683,
11,
597,
307,
411,
11,
286,
727,
1884,
1176,
42,
12683,
1507,
13,
51804,
51804
]
},
{
"avg_logprob": -0.37404739972457146,
"compression_ratio": 1.5743801652892562,
"end": 1044,
"id": 152,
"no_speech_prob": 0.00012728571891784668,
"seek": 102176,
"start": 1039.36,
"temperature": 0,
"text": " Or should we put it together? I don't think there's an indicator of all those fields. I know that there's",
"tokens": [
50364,
466,
257,
18350,
295,
479,
23728,
11,
295,
38024,
770,
2539,
11,
15756,
11427,
11,
293,
3009,
1176,
42,
12683,
293,
439,
50776,
50776,
1080,
2098,
293,
6422,
13,
1119,
456,
257,
665,
439,
28268,
278,
5934,
337,
341,
4992,
30,
51104,
51244,
1610,
820,
321,
829,
309,
1214,
30,
286,
500,
380,
519,
456,
311,
364,
16961,
295,
439,
729,
7909,
13,
286,
458,
300,
456,
311,
51476,
51476,
512,
479,
23728,
1507,
13,
286,
600,
829,
1214,
257,
5934,
1219,
611,
1176,
42,
12683,
11,
597,
307,
411,
11,
286,
727,
1884,
1176,
42,
12683,
1507,
13,
51804,
51804
]
},
{
"avg_logprob": -0.37404739972457146,
"compression_ratio": 1.5743801652892562,
"end": 1050.56,
"id": 153,
"no_speech_prob": 0.00012728571891784668,
"seek": 102176,
"start": 1044,
"temperature": 0,
"text": " some FHE stuff. I've put together a guide called also ZKML, which is like, I could create ZKML stuff.",
"tokens": [
50364,
466,
257,
18350,
295,
479,
23728,
11,
295,
38024,
770,
2539,
11,
15756,
11427,
11,
293,
3009,
1176,
42,
12683,
293,
439,
50776,
50776,
1080,
2098,
293,
6422,
13,
1119,
456,
257,
665,
439,
28268,
278,
5934,
337,
341,
4992,
30,
51104,
51244,
1610,
820,
321,
829,
309,
1214,
30,
286,
500,
380,
519,
456,
311,
364,
16961,
295,
439,
729,
7909,
13,
286,
458,
300,
456,
311,
51476,
51476,
512,
479,
23728,
1507,
13,
286,
600,
829,
1214,
257,
5934,
1219,
611,
1176,
42,
12683,
11,
597,
307,
411,
11,
286,
727,
1884,
1176,
42,
12683,
1507,
13,
51804,
51804
]
},
{
"avg_logprob": -0.18950300013765375,
"compression_ratio": 1.726530612244898,
"end": 1058.08,
"id": 154,
"no_speech_prob": 0.0000446803132945206,
"seek": 105056,
"start": 1050.56,
"temperature": 0,
"text": " I haven't seen anything about differential privacy. I haven't looked into that topic,",
"tokens": [
50364,
286,
2378,
380,
1612,
1340,
466,
15756,
11427,
13,
286,
2378,
380,
2956,
666,
300,
4829,
11,
50740,
50740,
457,
286,
478,
1238,
988,
300,
291,
393,
445,
1884,
257,
3211,
300,
445,
49866,
439,
295,
613,
3593,
1214,
13,
51040,
51128,
865,
11,
1392,
13,
286,
445,
2942,
257,
688,
295,
1880,
21267,
300,
291,
393,
689,
291,
2890,
51316,
51316,
732,
420,
1045,
295,
613,
958,
5125,
7943,
293,
291,
483,
5101,
512,
777,
51540,
51568,
777,
13759,
300,
291,
393,
380,
9021,
365,
2139,
295,
552,
13,
51800,
51800
]
},
{
"avg_logprob": -0.18950300013765375,
"compression_ratio": 1.726530612244898,
"end": 1064.08,
"id": 155,
"no_speech_prob": 0.0000446803132945206,
"seek": 105056,
"start": 1058.08,
"temperature": 0,
"text": " but I'm pretty sure that you can just create a doc that just encompasses all of these resources together.",
"tokens": [
50364,
286,
2378,
380,
1612,
1340,
466,
15756,
11427,
13,
286,
2378,
380,
2956,
666,
300,
4829,
11,
50740,
50740,
457,
286,
478,
1238,
988,
300,
291,
393,
445,
1884,
257,
3211,
300,
445,
49866,
439,
295,
613,
3593,
1214,
13,
51040,
51128,
865,
11,
1392,
13,
286,
445,
2942,
257,
688,
295,
1880,
21267,
300,
291,
393,
689,
291,
2890,
51316,
51316,
732,
420,
1045,
295,
613,
958,
5125,
7943,
293,
291,
483,
5101,
512,
777,
51540,
51568,
777,
13759,
300,
291,
393,
380,
9021,
365,
2139,
295,
552,
13,
51800,
51800
]
},
{
"avg_logprob": -0.18950300013765375,
"compression_ratio": 1.726530612244898,
"end": 1069.6,
"id": 156,
"no_speech_prob": 0.0000446803132945206,
"seek": 105056,
"start": 1065.84,
"temperature": 0,
"text": " Yeah, okay. I just created a lot of interesting combinations that you can where you mix",
"tokens": [
50364,
286,
2378,
380,
1612,
1340,
466,
15756,
11427,
13,
286,
2378,
380,
2956,
666,
300,
4829,
11,
50740,
50740,
457,
286,
478,
1238,
988,
300,
291,
393,
445,
1884,
257,
3211,
300,
445,
49866,
439,
295,
613,
3593,
1214,
13,
51040,
51128,
865,
11,
1392,
13,
286,
445,
2942,
257,
688,
295,
1880,
21267,
300,
291,
393,
689,
291,
2890,
51316,
51316,
732,
420,
1045,
295,
613,
958,
5125,
7943,
293,
291,
483,
5101,
512,
777,
51540,
51568,
777,
13759,
300,
291,
393,
380,
9021,
365,
2139,
295,
552,
13,
51800,
51800
]
},
{
"avg_logprob": -0.18950300013765375,
"compression_ratio": 1.726530612244898,
"end": 1074.08,
"id": 157,
"no_speech_prob": 0.0000446803132945206,
"seek": 105056,
"start": 1069.6,
"temperature": 0,
"text": " two or three of these next generation technologies and you get generally some new",
"tokens": [
50364,
286,
2378,
380,
1612,
1340,
466,
15756,
11427,
13,
286,
2378,
380,
2956,
666,
300,
4829,
11,
50740,
50740,
457,
286,
478,
1238,
988,
300,
291,
393,
445,
1884,
257,
3211,
300,
445,
49866,
439,
295,
613,
3593,
1214,
13,
51040,
51128,
865,
11,
1392,
13,
286,
445,
2942,
257,
688,
295,
1880,
21267,
300,
291,
393,
689,
291,
2890,
51316,
51316,
732,
420,
1045,
295,
613,
958,
5125,
7943,
293,
291,
483,
5101,
512,
777,
51540,
51568,
777,
13759,
300,
291,
393,
380,
9021,
365,
2139,
295,
552,
13,
51800,
51800
]
},
{
"avg_logprob": -0.18950300013765375,
"compression_ratio": 1.726530612244898,
"end": 1079.28,
"id": 158,
"no_speech_prob": 0.0000446803132945206,
"seek": 105056,
"start": 1074.6399999999999,
"temperature": 0,
"text": " new capability that you can't accomplish with either of them.",
"tokens": [
50364,
286,
2378,
380,
1612,
1340,
466,
15756,
11427,
13,
286,
2378,
380,
2956,
666,
300,
4829,
11,
50740,
50740,
457,
286,
478,
1238,
988,
300,
291,
393,
445,
1884,
257,
3211,
300,
445,
49866,
439,
295,
613,
3593,
1214,
13,
51040,
51128,
865,
11,
1392,
13,
286,
445,
2942,
257,
688,
295,
1880,
21267,
300,
291,
393,
689,
291,
2890,
51316,
51316,
732,
420,
1045,
295,
613,
958,
5125,
7943,
293,
291,
483,
5101,
512,
777,
51540,
51568,
777,
13759,
300,
291,
393,
380,
9021,
365,
2139,
295,
552,
13,
51800,
51800
]
},
{
"avg_logprob": -0.18539473695575065,
"compression_ratio": 1.686131386861314,
"end": 1083.04,
"id": 159,
"no_speech_prob": 0.00002111025241902098,
"seek": 107928,
"start": 1079.28,
"temperature": 0,
"text": " So I think there'll probably be a lot of interesting things that come out that combine",
"tokens": [
50364,
407,
286,
519,
456,
603,
1391,
312,
257,
688,
295,
1880,
721,
300,
808,
484,
300,
10432,
50552,
50552,
15756,
11427,
293,
1176,
42,
420,
479,
23728,
420,
2035,
13,
14129,
604,
732,
293,
456,
311,
512,
777,
50872,
50872,
777,
2445,
291,
393,
14722,
13,
467,
311,
534,
1627,
11,
457,
264,
28270,
19922,
291,
1884,
490,
341,
51176,
51212,
562,
291,
10432,
7943,
411,
341,
307,
1238,
955,
13,
1743,
1266,
11,
1360,
4860,
28270,
19922,
51540,
51540,
670,
264,
24903,
2564,
11,
597,
307,
1019,
13,
583,
10807,
366,
370,
2370,
293,
321,
393,
380,
3496,
309,
787,
322,
51848,
51848
]
},
{
"avg_logprob": -0.18539473695575065,
"compression_ratio": 1.686131386861314,
"end": 1089.44,
"id": 160,
"no_speech_prob": 0.00002111025241902098,
"seek": 107928,
"start": 1083.04,
"temperature": 0,
"text": " differential privacy and ZK or FHE or whatever. Pick any two and there's some new",
"tokens": [
50364,
407,
286,
519,
456,
603,
1391,
312,
257,
688,
295,
1880,
721,
300,
808,
484,
300,
10432,
50552,
50552,
15756,
11427,
293,
1176,
42,
420,
479,
23728,
420,
2035,
13,
14129,
604,
732,
293,
456,
311,
512,
777,
50872,
50872,
777,
2445,
291,
393,
14722,
13,
467,
311,
534,
1627,
11,
457,
264,
28270,
19922,
291,
1884,
490,
341,
51176,
51212,
562,
291,
10432,
7943,
411,
341,
307,
1238,
955,
13,
1743,
1266,
11,
1360,
4860,
28270,
19922,
51540,
51540,
670,
264,
24903,
2564,
11,
597,
307,
1019,
13,
583,
10807,
366,
370,
2370,
293,
321,
393,
380,
3496,
309,
787,
322,
51848,
51848
]
},
{
"avg_logprob": -0.18539473695575065,
"compression_ratio": 1.686131386861314,
"end": 1095.52,
"id": 161,
"no_speech_prob": 0.00002111025241902098,
"seek": 107928,
"start": 1089.44,
"temperature": 0,
"text": " new function you can compute. It's really cool, but the computational overhead you create from this",
"tokens": [
50364,
407,
286,
519,
456,
603,
1391,
312,
257,
688,
295,
1880,
721,
300,
808,
484,
300,
10432,
50552,
50552,
15756,
11427,
293,
1176,
42,
420,
479,
23728,
420,
2035,
13,
14129,
604,
732,
293,
456,
311,
512,
777,
50872,
50872,
777,
2445,
291,
393,
14722,
13,
467,
311,
534,
1627,
11,
457,
264,
28270,
19922,
291,
1884,
490,
341,
51176,
51212,
562,
291,
10432,
7943,
411,
341,
307,
1238,
955,
13,
1743,
1266,
11,
1360,
4860,
28270,
19922,
51540,
51540,
670,
264,
24903,
2564,
11,
597,
307,
1019,
13,
583,
10807,
366,
370,
2370,
293,
321,
393,
380,
3496,
309,
787,
322,
51848,
51848
]
},
{
"avg_logprob": -0.18539473695575065,
"compression_ratio": 1.686131386861314,
"end": 1102.8,
"id": 162,
"no_speech_prob": 0.00002111025241902098,
"seek": 107928,
"start": 1096.24,
"temperature": 0,
"text": " when you combine technologies like this is pretty big. Like 10,000 fold computational overhead",
"tokens": [
50364,
407,
286,
519,
456,
603,
1391,
312,
257,
688,
295,
1880,
721,
300,
808,
484,
300,
10432,
50552,
50552,
15756,
11427,
293,
1176,
42,
420,
479,
23728,
420,
2035,
13,
14129,
604,
732,
293,
456,
311,
512,
777,
50872,
50872,
777,
2445,
291,
393,
14722,
13,
467,
311,
534,
1627,
11,
457,
264,
28270,
19922,
291,
1884,
490,
341,
51176,
51212,
562,
291,
10432,
7943,
411,
341,
307,
1238,
955,
13,
1743,
1266,
11,
1360,
4860,
28270,
19922,
51540,
51540,
670,
264,
24903,
2564,
11,
597,
307,
1019,
13,
583,
10807,
366,
370,
2370,
293,
321,
393,
380,
3496,
309,
787,
322,
51848,
51848
]
},
{
"avg_logprob": -0.18539473695575065,
"compression_ratio": 1.686131386861314,
"end": 1108.96,
"id": 163,
"no_speech_prob": 0.00002111025241902098,
"seek": 107928,
"start": 1102.8,
"temperature": 0,
"text": " over the computation itself, which is fun. But computers are so fast and we can't spend it only on",
"tokens": [
50364,
407,
286,
519,
456,
603,
1391,
312,
257,
688,
295,
1880,
721,
300,
808,
484,
300,
10432,
50552,
50552,
15756,
11427,
293,
1176,
42,
420,
479,
23728,
420,
2035,
13,
14129,
604,
732,
293,
456,
311,
512,
777,
50872,
50872,
777,
2445,
291,
393,
14722,
13,
467,
311,
534,
1627,
11,
457,
264,
28270,
19922,
291,
1884,
490,
341,
51176,
51212,
562,
291,
10432,
7943,
411,
341,
307,
1238,
955,
13,
1743,
1266,
11,
1360,
4860,
28270,
19922,
51540,
51540,
670,
264,
24903,
2564,
11,
597,
307,
1019,
13,
583,
10807,
366,
370,
2370,
293,
321,
393,
380,
3496,
309,
787,
322,
51848,
51848
]
},
{
"avg_logprob": -0.17822948031955296,
"compression_ratio": 1.509009009009009,
"end": 1115.52,
"id": 164,
"no_speech_prob": 0.00014648973592557013,
"seek": 110896,
"start": 1108.96,
"temperature": 0,
"text": " your JavaScript animation. Have you heard that the actual JavaScript ecosystem is actually",
"tokens": [
50364,
428,
15778,
9603,
13,
3560,
291,
2198,
300,
264,
3539,
15778,
11311,
307,
767,
50692,
50692,
1242,
1101,
586,
30,
509,
434,
1228,
411,
34952,
281,
352,
2844,
13,
467,
311,
445,
411,
777,
1507,
13,
467,
311,
534,
1019,
13,
50980,
50980,
467,
311,
1242,
1101,
13,
467,
1143,
281,
312,
5324,
13,
440,
14116,
6589,
574,
322,
9116,
311,
1851,
13,
51236,
51404,
3146,
264,
636,
11,
9114,
27377,
11,
630,
291,
483,
281,
751,
281,
264,
2146,
490,
7772,
12683,
13,
1301,
30,
51656,
51772
]
},
{
"avg_logprob": -0.17822948031955296,
"compression_ratio": 1.509009009009009,
"end": 1121.28,
"id": 165,
"no_speech_prob": 0.00014648973592557013,
"seek": 110896,
"start": 1115.52,
"temperature": 0,
"text": " getting better now? You're using like Rust to go pack. It's just like new stuff. It's really fun.",
"tokens": [
50364,
428,
15778,
9603,
13,
3560,
291,
2198,
300,
264,
3539,
15778,
11311,
307,
767,
50692,
50692,
1242,
1101,
586,
30,
509,
434,
1228,
411,
34952,
281,
352,
2844,
13,
467,
311,
445,
411,
777,
1507,
13,
467,
311,
534,
1019,
13,
50980,
50980,
467,
311,
1242,
1101,
13,
467,
1143,
281,
312,
5324,
13,
440,
14116,
6589,
574,
322,
9116,
311,
1851,
13,
51236,
51404,
3146,
264,
636,
11,
9114,
27377,
11,
630,
291,
483,
281,
751,
281,
264,
2146,
490,
7772,
12683,
13,
1301,
30,
51656,
51772
]
},
{
"avg_logprob": -0.17822948031955296,
"compression_ratio": 1.509009009009009,
"end": 1126.4,
"id": 166,
"no_speech_prob": 0.00014648973592557013,
"seek": 110896,
"start": 1121.28,
"temperature": 0,
"text": " It's getting better. It used to be worse. The disgusted look on Ryan's face.",
"tokens": [
50364,
428,
15778,
9603,
13,
3560,
291,
2198,
300,
264,
3539,
15778,
11311,
307,
767,
50692,
50692,
1242,
1101,
586,
30,
509,
434,
1228,
411,
34952,
281,
352,
2844,
13,
467,
311,
445,
411,
777,
1507,
13,
467,
311,
534,
1019,
13,
50980,
50980,
467,
311,
1242,
1101,
13,
467,
1143,
281,
312,
5324,
13,
440,
14116,
6589,
574,
322,
9116,
311,
1851,
13,
51236,
51404,
3146,
264,
636,
11,
9114,
27377,
11,
630,
291,
483,
281,
751,
281,
264,
2146,
490,
7772,
12683,
13,
1301,
30,
51656,
51772
]
},
{
"avg_logprob": -0.17822948031955296,
"compression_ratio": 1.509009009009009,
"end": 1134.8,
"id": 167,
"no_speech_prob": 0.00014648973592557013,
"seek": 110896,
"start": 1129.76,
"temperature": 0,
"text": " By the way, DC builder, did you get to talk to the guy from FedML.ai?",
"tokens": [
50364,
428,
15778,
9603,
13,
3560,
291,
2198,
300,
264,
3539,
15778,
11311,
307,
767,
50692,
50692,
1242,
1101,
586,
30,
509,
434,
1228,
411,
34952,
281,
352,
2844,
13,
467,
311,
445,
411,
777,
1507,
13,
467,
311,
534,
1019,
13,
50980,
50980,
467,
311,
1242,
1101,
13,
467,
1143,
281,
312,
5324,
13,
440,
14116,
6589,
574,
322,
9116,
311,
1851,
13,
51236,
51404,
3146,
264,
636,
11,
9114,
27377,
11,
630,
291,
483,
281,
751,
281,
264,
2146,
490,
7772,
12683,
13,
1301,
30,
51656,
51772
]
},
{
"avg_logprob": -0.30943854836856616,
"compression_ratio": 1.477832512315271,
"end": 1140.72,
"id": 168,
"no_speech_prob": 0.0002912460477091372,
"seek": 113480,
"start": 1134.8,
"temperature": 0,
"text": " FedML, I don't think I've heard about that.",
"tokens": [
50364,
7772,
12683,
11,
286,
500,
380,
519,
286,
600,
2198,
466,
300,
13,
50660,
50660,
286,
1194,
286,
4582,
291,
1074,
13,
865,
11,
286,
9437,
309,
322,
264,
5081,
13,
50920,
50920,
865,
11,
3231,
13,
50980,
50980,
407,
309,
311,
257,
3663,
337,
884,
38024,
770,
3479,
2539,
13,
8537,
11,
436,
2942,
257,
3840,
295,
51264,
51264,
2886,
38970,
13,
865,
13,
400,
436,
434,
611,
11,
436,
434,
611,
11225,
257,
1238,
955,
8871,
420,
2638,
316,
13,
51608
]
},
{
"avg_logprob": -0.30943854836856616,
"compression_ratio": 1.477832512315271,
"end": 1145.9199999999998,
"id": 169,
"no_speech_prob": 0.0002912460477091372,
"seek": 113480,
"start": 1140.72,
"temperature": 0,
"text": " I thought I connected you guys. Yeah, I posted it on the chat.",
"tokens": [
50364,
7772,
12683,
11,
286,
500,
380,
519,
286,
600,
2198,
466,
300,
13,
50660,
50660,
286,
1194,
286,
4582,
291,
1074,
13,
865,
11,
286,
9437,
309,
322,
264,
5081,
13,
50920,
50920,
865,
11,
3231,
13,
50980,
50980,
407,
309,
311,
257,
3663,
337,
884,
38024,
770,
3479,
2539,
13,
8537,
11,
436,
2942,
257,
3840,
295,
51264,
51264,
2886,
38970,
13,
865,
13,
400,
436,
434,
611,
11,
436,
434,
611,
11225,
257,
1238,
955,
8871,
420,
2638,
316,
13,
51608
]
},
{
"avg_logprob": -0.30943854836856616,
"compression_ratio": 1.477832512315271,
"end": 1147.12,
"id": 170,
"no_speech_prob": 0.0002912460477091372,
"seek": 113480,
"start": 1145.9199999999998,
"temperature": 0,
"text": " Yeah, thanks.",
"tokens": [
50364,
7772,
12683,
11,
286,
500,
380,
519,
286,
600,
2198,
466,
300,
13,
50660,
50660,
286,
1194,
286,
4582,
291,
1074,
13,
865,
11,
286,
9437,
309,
322,
264,
5081,
13,
50920,
50920,
865,
11,
3231,
13,
50980,
50980,
407,
309,
311,
257,
3663,
337,
884,
38024,
770,
3479,
2539,
13,
8537,
11,
436,
2942,
257,
3840,
295,
51264,
51264,
2886,
38970,
13,
865,
13,
400,
436,
434,
611,
11,
436,
434,
611,
11225,
257,
1238,
955,
8871,
420,
2638,
316,
13,
51608
]
},
{
"avg_logprob": -0.30943854836856616,
"compression_ratio": 1.477832512315271,
"end": 1152.8,
"id": 171,
"no_speech_prob": 0.0002912460477091372,
"seek": 113480,
"start": 1147.12,
"temperature": 0,
"text": " So it's a platform for doing federated machine learning. Basically, they created a bunch of",
"tokens": [
50364,
7772,
12683,
11,
286,
500,
380,
519,
286,
600,
2198,
466,
300,
13,
50660,
50660,
286,
1194,
286,
4582,
291,
1074,
13,
865,
11,
286,
9437,
309,
322,
264,
5081,
13,
50920,
50920,
865,
11,
3231,
13,
50980,
50980,
407,
309,
311,
257,
3663,
337,
884,
38024,
770,
3479,
2539,
13,
8537,
11,
436,
2942,
257,
3840,
295,
51264,
51264,
2886,
38970,
13,
865,
13,
400,
436,
434,
611,
11,
436,
434,
611,
11225,
257,
1238,
955,
8871,
420,
2638,
316,
13,
51608
]
},
{
"avg_logprob": -0.2645183423670327,
"compression_ratio": 1.282442748091603,
"end": 1168.8799999999999,
"id": 172,
"no_speech_prob": 0.00012727457215078175,
"seek": 115280,
"start": 1152.8,
"temperature": 0,
"text": " primitives. Yeah. And they're also racing a pretty big seed or series A.",
"tokens": [
50364,
2886,
38970,
13,
865,
13,
400,
436,
434,
611,
12553,
257,
1238,
955,
8871,
420,
2638,
316,
13,
51168,
51168,
492,
393,
751,
466,
11427,
1310,
13,
286,
3264,
512,
18106,
9130,
466,
2416,
2856,
5245,
294,
51484,
51484
]
},
{
"avg_logprob": -0.2645183423670327,
"compression_ratio": 1.282442748091603,
"end": 1175.2,
"id": 173,
"no_speech_prob": 0.00012727457215078175,
"seek": 115280,
"start": 1168.8799999999999,
"temperature": 0,
"text": " We can talk about privacy maybe. I learned some terrifying facts about large language models in",
"tokens": [
50364,
2886,
38970,
13,
865,
13,
400,
436,
434,
611,
12553,
257,
1238,
955,
8871,
420,
2638,
316,
13,
51168,
51168,
492,
393,
751,
466,
11427,
1310,
13,
286,
3264,
512,
18106,
9130,
466,
2416,
2856,
5245,
294,
51484,
51484
]
},
{
"avg_logprob": -0.26344389664499385,
"compression_ratio": 1.4619289340101522,
"end": 1186,
"id": 174,
"no_speech_prob": 0.00008347872062586248,
"seek": 117520,
"start": 1175.2,
"temperature": 0,
"text": " this ZKL knowledge podcast, with Florian Tramier, about how you get basically like private data",
"tokens": [
50364,
341,
1176,
42,
43,
3601,
7367,
11,
365,
8328,
952,
1765,
335,
811,
11,
466,
577,
291,
483,
1936,
411,
4551,
1412,
50904,
50904,
16979,
31779,
807,
26039,
51,
12,
18,
420,
26039,
51,
12,
17,
300,
286,
632,
572,
1558,
466,
13,
1449,
411,
1580,
15715,
51260,
51260,
9437,
456,
420,
658,
11,
286,
2041,
11,
360,
30569,
292,
2950,
293,
550,
264,
3097,
1412,
445,
6183,
309,
493,
293,
264,
51712,
51712
]
},
{
"avg_logprob": -0.26344389664499385,
"compression_ratio": 1.4619289340101522,
"end": 1193.1200000000001,
"id": 175,
"no_speech_prob": 0.00008347872062586248,
"seek": 117520,
"start": 1186,
"temperature": 0,
"text": " randomly leaked through GPT-3 or GPT-2 that I had no idea about. Just like someone accidentally",
"tokens": [
50364,
341,
1176,
42,
43,
3601,
7367,
11,
365,
8328,
952,
1765,
335,
811,
11,
466,
577,
291,
483,
1936,
411,
4551,
1412,
50904,
50904,
16979,
31779,
807,
26039,
51,
12,
18,
420,
26039,
51,
12,
17,
300,
286,
632,
572,
1558,
466,
13,
1449,
411,
1580,
15715,
51260,
51260,
9437,
456,
420,
658,
11,
286,
2041,
11,
360,
30569,
292,
2950,
293,
550,
264,
3097,
1412,
445,
6183,
309,
493,
293,
264,
51712,
51712
]
},
{
"avg_logprob": -0.26344389664499385,
"compression_ratio": 1.4619289340101522,
"end": 1202.16,
"id": 176,
"no_speech_prob": 0.00008347872062586248,
"seek": 117520,
"start": 1193.1200000000001,
"temperature": 0,
"text": " posted there or got, I guess, doxxed online and then the training data just picked it up and the",
"tokens": [
50364,
341,
1176,
42,
43,
3601,
7367,
11,
365,
8328,
952,
1765,
335,
811,
11,
466,
577,
291,
483,
1936,
411,
4551,
1412,
50904,
50904,
16979,
31779,
807,
26039,
51,
12,
18,
420,
26039,
51,
12,
17,
300,
286,
632,
572,
1558,
466,
13,
1449,
411,
1580,
15715,
51260,
51260,
9437,
456,
420,
658,
11,
286,
2041,
11,
360,
30569,
292,
2950,
293,
550,
264,
3097,
1412,
445,
6183,
309,
493,
293,
264,
51712,
51712
]
},
{
"avg_logprob": -0.17755067503297484,
"compression_ratio": 1.4097560975609755,
"end": 1209.92,
"id": 177,
"no_speech_prob": 0.000029770026230835356,
"seek": 120216,
"start": 1202.16,
"temperature": 0,
"text": " model didn't transform it. So some people just included in the output of GPT-2 with their full",
"tokens": [
50364,
2316,
994,
380,
4088,
309,
13,
407,
512,
561,
445,
5556,
294,
264,
5598,
295,
26039,
51,
12,
17,
365,
641,
1577,
50752,
50752,
1412,
13,
286,
2041,
286,
393,
1710,
13,
440,
636,
613,
11,
412,
1935,
264,
2440,
733,
295,
411,
45702,
45,
441,
6840,
44,
12,
6032,
5245,
589,
51352,
51352,
307,
300,
337,
1184,
4846,
14862,
11,
291,
366,
2745,
5145,
2837,
295,
264,
1785,
295,
264,
8174,
420,
51676,
51676
]
},
{
"avg_logprob": -0.17755067503297484,
"compression_ratio": 1.4097560975609755,
"end": 1221.92,
"id": 178,
"no_speech_prob": 0.000029770026230835356,
"seek": 120216,
"start": 1209.92,
"temperature": 0,
"text": " data. I guess I can speak. The way these, at least the early kind of like RNN LSTM-based models work",
"tokens": [
50364,
2316,
994,
380,
4088,
309,
13,
407,
512,
561,
445,
5556,
294,
264,
5598,
295,
26039,
51,
12,
17,
365,
641,
1577,
50752,
50752,
1412,
13,
286,
2041,
286,
393,
1710,
13,
440,
636,
613,
11,
412,
1935,
264,
2440,
733,
295,
411,
45702,
45,
441,
6840,
44,
12,
6032,
5245,
589,
51352,
51352,
307,
300,
337,
1184,
4846,
14862,
11,
291,
366,
2745,
5145,
2837,
295,
264,
1785,
295,
264,
8174,
420,
51676,
51676
]
},
{
"avg_logprob": -0.17755067503297484,
"compression_ratio": 1.4097560975609755,
"end": 1228.4,
"id": 179,
"no_speech_prob": 0.000029770026230835356,
"seek": 120216,
"start": 1221.92,
"temperature": 0,
"text": " is that for each input token, you are obviously keeping track of the state of the sentence or",
"tokens": [
50364,
2316,
994,
380,
4088,
309,
13,
407,
512,
561,
445,
5556,
294,
264,
5598,
295,
26039,
51,
12,
17,
365,
641,
1577,
50752,
50752,
1412,
13,
286,
2041,
286,
393,
1710,
13,
440,
636,
613,
11,
412,
1935,
264,
2440,
733,
295,
411,
45702,
45,
441,
6840,
44,
12,
6032,
5245,
589,
51352,
51352,
307,
300,
337,
1184,
4846,
14862,
11,
291,
366,
2745,
5145,
2837,
295,
264,
1785,
295,
264,
8174,
420,
51676,
51676
]
},
{
"avg_logprob": -0.08884678063569246,
"compression_ratio": 1.7454545454545454,
"end": 1234.88,
"id": 180,
"no_speech_prob": 0.000020783945728908293,
"seek": 122840,
"start": 1228.4,
"temperature": 0,
"text": " your context. But basically, you are trying to output the greatest probability next token based",
"tokens": [
50364,
428,
4319,
13,
583,
1936,
11,
291,
366,
1382,
281,
5598,
264,
6636,
8482,
958,
14862,
2361,
50688,
50688,
322,
428,
4319,
370,
1400,
13,
400,
291,
393,
829,
3202,
670,
341,
498,
291,
362,
787,
3202,
300,
291,
362,
50940,
50940,
257,
31782,
13,
583,
264,
1558,
510,
307,
300,
498,
456,
366,
588,
11,
588,
3845,
22978,
294,
264,
1412,
992,
11,
51268,
51268,
293,
498,
428,
411,
2316,
10139,
1860,
11,
370,
411,
428,
7633,
1785,
10139,
1860,
307,
534,
1090,
11,
51564,
51564,
550,
437,
5314,
493,
2737,
307,
264,
2316,
1596,
3736,
445,
10560,
5660,
570,
309,
1619,
11,
51772,
51772
]
},
{
"avg_logprob": -0.08884678063569246,
"compression_ratio": 1.7454545454545454,
"end": 1239.92,
"id": 181,
"no_speech_prob": 0.000020783945728908293,
"seek": 122840,
"start": 1234.88,
"temperature": 0,
"text": " on your context so far. And you can put attention over this if you have only attention that you have",
"tokens": [
50364,
428,
4319,
13,
583,
1936,
11,
291,
366,
1382,
281,
5598,
264,
6636,
8482,
958,
14862,
2361,
50688,
50688,
322,
428,
4319,
370,
1400,
13,
400,
291,
393,
829,
3202,
670,
341,
498,
291,
362,
787,
3202,
300,
291,
362,
50940,
50940,
257,
31782,
13,
583,
264,
1558,
510,
307,
300,
498,
456,
366,
588,
11,
588,
3845,
22978,
294,
264,
1412,
992,
11,
51268,
51268,
293,
498,
428,
411,
2316,
10139,
1860,
11,
370,
411,
428,
7633,
1785,
10139,
1860,
307,
534,
1090,
11,
51564,
51564,
550,
437,
5314,
493,
2737,
307,
264,
2316,
1596,
3736,
445,
10560,
5660,
570,
309,
1619,
11,
51772,
51772
]
},
{
"avg_logprob": -0.08884678063569246,
"compression_ratio": 1.7454545454545454,
"end": 1246.48,
"id": 182,
"no_speech_prob": 0.000020783945728908293,
"seek": 122840,
"start": 1239.92,
"temperature": 0,
"text": " a transformer. But the idea here is that if there are very, very unique sequences in the data set,",
"tokens": [
50364,
428,
4319,
13,
583,
1936,
11,
291,
366,
1382,
281,
5598,
264,
6636,
8482,
958,
14862,
2361,
50688,
50688,
322,
428,
4319,
370,
1400,
13,
400,
291,
393,
829,
3202,
670,
341,
498,
291,
362,
787,
3202,
300,
291,
362,
50940,
50940,
257,
31782,
13,
583,
264,
1558,
510,
307,
300,
498,
456,
366,
588,
11,
588,
3845,
22978,
294,
264,
1412,
992,
11,
51268,
51268,
293,
498,
428,
411,
2316,
10139,
1860,
11,
370,
411,
428,
7633,
1785,
10139,
1860,
307,
534,
1090,
11,
51564,
51564,
550,
437,
5314,
493,
2737,
307,
264,
2316,
1596,
3736,
445,
10560,
5660,
570,
309,
1619,
11,
51772,
51772
]
},
{
"avg_logprob": -0.08884678063569246,
"compression_ratio": 1.7454545454545454,
"end": 1252.4,
"id": 183,
"no_speech_prob": 0.000020783945728908293,
"seek": 122840,
"start": 1246.48,
"temperature": 0,
"text": " and if your like model dimensionality, so like your hidden state dimensionality is really high,",
"tokens": [
50364,
428,
4319,
13,
583,
1936,
11,
291,
366,
1382,
281,
5598,
264,
6636,
8482,
958,
14862,
2361,
50688,
50688,
322,
428,
4319,
370,
1400,
13,
400,
291,
393,
829,
3202,
670,
341,
498,
291,
362,
787,
3202,
300,
291,
362,
50940,
50940,
257,
31782,
13,
583,
264,
1558,
510,
307,
300,
498,
456,
366,
588,
11,
588,
3845,
22978,
294,
264,
1412,
992,
11,
51268,
51268,
293,
498,
428,
411,
2316,
10139,
1860,
11,
370,
411,
428,
7633,
1785,
10139,
1860,
307,
534,
1090,
11,
51564,
51564,
550,
437,
5314,
493,
2737,
307,
264,
2316,
1596,
3736,
445,
10560,
5660,
570,
309,
1619,
11,
51772,
51772
]
},
{
"avg_logprob": -0.08884678063569246,
"compression_ratio": 1.7454545454545454,
"end": 1256.5600000000002,
"id": 184,
"no_speech_prob": 0.000020783945728908293,
"seek": 122840,
"start": 1252.4,
"temperature": 0,
"text": " then what ends up happening is the model quite literally just memorizes because it says,",
"tokens": [
50364,
428,
4319,
13,
583,
1936,
11,
291,
366,
1382,
281,
5598,
264,
6636,
8482,
958,
14862,
2361,
50688,
50688,
322,
428,
4319,
370,
1400,
13,
400,
291,
393,
829,
3202,
670,
341,
498,
291,
362,
787,
3202,
300,
291,
362,
50940,
50940,
257,
31782,
13,
583,
264,
1558,
510,
307,
300,
498,
456,
366,
588,
11,
588,
3845,
22978,
294,
264,
1412,
992,
11,
51268,
51268,
293,
498,
428,
411,
2316,
10139,
1860,
11,
370,
411,
428,
7633,
1785,
10139,
1860,
307,
534,
1090,
11,
51564,
51564,
550,
437,
5314,
493,
2737,
307,
264,
2316,
1596,
3736,
445,
10560,
5660,
570,
309,
1619,
11,
51772,
51772
]
},
{
"avg_logprob": -0.07270262791560246,
"compression_ratio": 1.6376811594202898,
"end": 1261.36,
"id": 185,
"no_speech_prob": 0.0000483239273307845,
"seek": 125656,
"start": 1256.56,
"temperature": 0,
"text": " yeah, after this particular sequence of tokens, the only possible next token that I've seen in",
"tokens": [
50364,
1338,
11,
934,
341,
1729,
8310,
295,
22667,
11,
264,
787,
1944,
958,
14862,
300,
286,
600,
1612,
294,
50604,
50604,
264,
1379,
8149,
11,
597,
307,
406,
516,
281,
976,
385,
13785,
4470,
1936,
11,
307,
341,
958,
14862,
13,
50860,
50860,
400,
370,
309,
445,
637,
1208,
484,
300,
46677,
8310,
13,
639,
733,
295,
1154,
2544,
281,
312,
6264,
51204,
51204,
26387,
281,
445,
1090,
18795,
2316,
33358,
294,
2674,
13,
286,
519,
456,
311,
51496,
51496,
668,
512,
589,
1096,
322,
411,
1382,
281,
360,
13956,
1287,
293,
411,
3126,
29905,
322,
11,
286,
2041,
11,
51796,
51796
]
},
{
"avg_logprob": -0.07270262791560246,
"compression_ratio": 1.6376811594202898,
"end": 1266.48,
"id": 186,
"no_speech_prob": 0.0000483239273307845,
"seek": 125656,
"start": 1261.36,
"temperature": 0,
"text": " the whole database, which is not going to give me infinite loss basically, is this next token.",
"tokens": [
50364,
1338,
11,
934,
341,
1729,
8310,
295,
22667,
11,
264,
787,
1944,
958,
14862,
300,
286,
600,
1612,
294,
50604,
50604,
264,
1379,
8149,
11,
597,
307,
406,
516,
281,
976,
385,
13785,
4470,
1936,
11,
307,
341,
958,
14862,
13,
50860,
50860,
400,
370,
309,
445,
637,
1208,
484,
300,
46677,
8310,
13,
639,
733,
295,
1154,
2544,
281,
312,
6264,
51204,
51204,
26387,
281,
445,
1090,
18795,
2316,
33358,
294,
2674,
13,
286,
519,
456,
311,
51496,
51496,
668,
512,
589,
1096,
322,
411,
1382,
281,
360,
13956,
1287,
293,
411,
3126,
29905,
322,
11,
286,
2041,
11,
51796,
51796
]
},
{
"avg_logprob": -0.07270262791560246,
"compression_ratio": 1.6376811594202898,
"end": 1273.36,
"id": 187,
"no_speech_prob": 0.0000483239273307845,
"seek": 125656,
"start": 1266.48,
"temperature": 0,
"text": " And so it just spits out that memorized sequence. This kind of problem seems to be possibly",
"tokens": [
50364,
1338,
11,
934,
341,
1729,
8310,
295,
22667,
11,
264,
787,
1944,
958,
14862,
300,
286,
600,
1612,
294,
50604,
50604,
264,
1379,
8149,
11,
597,
307,
406,
516,
281,
976,
385,
13785,
4470,
1936,
11,
307,
341,
958,
14862,
13,
50860,
50860,
400,
370,
309,
445,
637,
1208,
484,
300,
46677,
8310,
13,
639,
733,
295,
1154,
2544,
281,
312,
6264,
51204,
51204,
26387,
281,
445,
1090,
18795,
2316,
33358,
294,
2674,
13,
286,
519,
456,
311,
51496,
51496,
668,
512,
589,
1096,
322,
411,
1382,
281,
360,
13956,
1287,
293,
411,
3126,
29905,
322,
11,
286,
2041,
11,
51796,
51796
]
},
{
"avg_logprob": -0.07270262791560246,
"compression_ratio": 1.6376811594202898,
"end": 1279.2,
"id": 188,
"no_speech_prob": 0.0000483239273307845,
"seek": 125656,
"start": 1273.36,
"temperature": 0,
"text": " inherent to just high dimensional model representations in general. I think there's",
"tokens": [
50364,
1338,
11,
934,
341,
1729,
8310,
295,
22667,
11,
264,
787,
1944,
958,
14862,
300,
286,
600,
1612,
294,
50604,
50604,
264,
1379,
8149,
11,
597,
307,
406,
516,
281,
976,
385,
13785,
4470,
1936,
11,
307,
341,
958,
14862,
13,
50860,
50860,
400,
370,
309,
445,
637,
1208,
484,
300,
46677,
8310,
13,
639,
733,
295,
1154,
2544,
281,
312,
6264,
51204,
51204,
26387,
281,
445,
1090,
18795,
2316,
33358,
294,
2674,
13,
286,
519,
456,
311,
51496,
51496,
668,
512,
589,
1096,
322,
411,
1382,
281,
360,
13956,
1287,
293,
411,
3126,
29905,
322,
11,
286,
2041,
11,
51796,
51796
]
},
{
"avg_logprob": -0.07270262791560246,
"compression_ratio": 1.6376811594202898,
"end": 1285.2,
"id": 189,
"no_speech_prob": 0.0000483239273307845,
"seek": 125656,
"start": 1279.2,
"temperature": 0,
"text": " been some work done on like trying to do robustness and like lower bounds on, I guess,",
"tokens": [
50364,
1338,
11,
934,
341,
1729,
8310,
295,
22667,
11,
264,
787,
1944,
958,
14862,
300,
286,
600,
1612,
294,
50604,
50604,
264,
1379,
8149,
11,
597,
307,
406,
516,
281,
976,
385,
13785,
4470,
1936,
11,
307,
341,
958,
14862,
13,
50860,
50860,
400,
370,
309,
445,
637,
1208,
484,
300,
46677,
8310,
13,
639,
733,
295,
1154,
2544,
281,
312,
6264,
51204,
51204,
26387,
281,
445,
1090,
18795,
2316,
33358,
294,
2674,
13,
286,
519,
456,
311,
51496,
51496,
668,
512,
589,
1096,
322,
411,
1382,
281,
360,
13956,
1287,
293,
411,
3126,
29905,
322,
11,
286,
2041,
11,
51796,
51796
]
},
{
"avg_logprob": -0.08634600593048392,
"compression_ratio": 1.6974169741697418,
"end": 1290.0800000000002,
"id": 190,
"no_speech_prob": 0.00011234792327741161,
"seek": 128520,
"start": 1285.2,
"temperature": 0,
"text": " like error or something like that. I'm not too familiar with the literature, but it does seem",
"tokens": [
50364,
411,
6713,
420,
746,
411,
300,
13,
286,
478,
406,
886,
4963,
365,
264,
10394,
11,
457,
309,
775,
1643,
50608,
50608,
411,
337,
613,
411,
733,
295,
15957,
3142,
3997,
10290,
5245,
11,
10560,
2144,
1062,
312,
411,
50912,
50912,
364,
26387,
1154,
13,
400,
611,
456,
311,
264,
1379,
2734,
365,
411,
7083,
1840,
11,
558,
11,
300,
8328,
952,
390,
51176,
51176,
1417,
466,
11,
294,
597,
291,
393,
1920,
411,
31499,
2316,
17443,
337,
4356,
5245,
51404,
51404,
538,
445,
7083,
1840,
264,
2316,
1547,
1413,
498,
291,
458,
264,
3877,
295,
264,
2316,
13,
51568,
51804
]
},
{
"avg_logprob": -0.08634600593048392,
"compression_ratio": 1.6974169741697418,
"end": 1296.16,
"id": 191,
"no_speech_prob": 0.00011234792327741161,
"seek": 128520,
"start": 1290.0800000000002,
"temperature": 0,
"text": " like for these like kind of deterministic complex representation models, memorization might be like",
"tokens": [
50364,
411,
6713,
420,
746,
411,
300,
13,
286,
478,
406,
886,
4963,
365,
264,
10394,
11,
457,
309,
775,
1643,
50608,
50608,
411,
337,
613,
411,
733,
295,
15957,
3142,
3997,
10290,
5245,
11,
10560,
2144,
1062,
312,
411,
50912,
50912,
364,
26387,
1154,
13,
400,
611,
456,
311,
264,
1379,
2734,
365,
411,
7083,
1840,
11,
558,
11,
300,
8328,
952,
390,
51176,
51176,
1417,
466,
11,
294,
597,
291,
393,
1920,
411,
31499,
2316,
17443,
337,
4356,
5245,
51404,
51404,
538,
445,
7083,
1840,
264,
2316,
1547,
1413,
498,
291,
458,
264,
3877,
295,
264,
2316,
13,
51568,
51804
]
},
{
"avg_logprob": -0.08634600593048392,
"compression_ratio": 1.6974169741697418,
"end": 1301.44,
"id": 192,
"no_speech_prob": 0.00011234792327741161,
"seek": 128520,
"start": 1296.16,
"temperature": 0,
"text": " an inherent problem. And also there's the whole issue with like querying, right, that Florian was",
"tokens": [
50364,
411,
6713,
420,
746,
411,
300,
13,
286,
478,
406,
886,
4963,
365,
264,
10394,
11,
457,
309,
775,
1643,
50608,
50608,
411,
337,
613,
411,
733,
295,
15957,
3142,
3997,
10290,
5245,
11,
10560,
2144,
1062,
312,
411,
50912,
50912,
364,
26387,
1154,
13,
400,
611,
456,
311,
264,
1379,
2734,
365,
411,
7083,
1840,
11,
558,
11,
300,
8328,
952,
390,
51176,
51176,
1417,
466,
11,
294,
597,
291,
393,
1920,
411,
31499,
2316,
17443,
337,
4356,
5245,
51404,
51404,
538,
445,
7083,
1840,
264,
2316,
1547,
1413,
498,
291,
458,
264,
3877,
295,
264,
2316,
13,
51568,
51804
]
},
{
"avg_logprob": -0.08634600593048392,
"compression_ratio": 1.6974169741697418,
"end": 1306,
"id": 193,
"no_speech_prob": 0.00011234792327741161,
"seek": 128520,
"start": 1301.44,
"temperature": 0,
"text": " talking about, in which you can almost like reconstruct model weights for smaller models",
"tokens": [
50364,
411,
6713,
420,
746,
411,
300,
13,
286,
478,
406,
886,
4963,
365,
264,
10394,
11,
457,
309,
775,
1643,
50608,
50608,
411,
337,
613,
411,
733,
295,
15957,
3142,
3997,
10290,
5245,
11,
10560,
2144,
1062,
312,
411,
50912,
50912,
364,
26387,
1154,
13,
400,
611,
456,
311,
264,
1379,
2734,
365,
411,
7083,
1840,
11,
558,
11,
300,
8328,
952,
390,
51176,
51176,
1417,
466,
11,
294,
597,
291,
393,
1920,
411,
31499,
2316,
17443,
337,
4356,
5245,
51404,
51404,
538,
445,
7083,
1840,
264,
2316,
1547,
1413,
498,
291,
458,
264,
3877,
295,
264,
2316,
13,
51568,
51804
]
},
{
"avg_logprob": -0.08634600593048392,
"compression_ratio": 1.6974169741697418,
"end": 1309.28,
"id": 194,
"no_speech_prob": 0.00011234792327741161,
"seek": 128520,
"start": 1306,
"temperature": 0,
"text": " by just querying the model enough times if you know the structure of the model.",
"tokens": [
50364,
411,
6713,
420,
746,
411,
300,
13,
286,
478,
406,
886,
4963,
365,
264,
10394,
11,
457,
309,
775,
1643,
50608,
50608,
411,
337,
613,
411,
733,
295,
15957,
3142,
3997,
10290,
5245,
11,
10560,
2144,
1062,
312,
411,
50912,
50912,
364,
26387,
1154,
13,
400,
611,
456,
311,
264,
1379,
2734,
365,
411,
7083,
1840,
11,
558,
11,
300,
8328,
952,
390,
51176,
51176,
1417,
466,
11,
294,
597,
291,
393,
1920,
411,
31499,
2316,
17443,
337,
4356,
5245,
51404,
51404,
538,
445,
7083,
1840,
264,
2316,
1547,
1413,
498,
291,
458,
264,
3877,
295,
264,
2316,
13,
51568,
51804
]
},
{
"avg_logprob": -0.26901139153374565,
"compression_ratio": 1.3588235294117648,
"end": 1317.52,
"id": 195,
"no_speech_prob": 0.00026941500254906714,
"seek": 130928,
"start": 1309.28,
"temperature": 0,
"text": " All right, thanks. That was very helpful. I had no idea how it works in particular,",
"tokens": [
50364,
1057,
558,
11,
3231,
13,
663,
390,
588,
4961,
13,
286,
632,
572,
1558,
577,
309,
1985,
294,
1729,
11,
50776,
50776,
457,
300,
390,
9045,
4745,
13,
50876,
51276,
27697,
286,
727,
652,
1101,
4337,
391,
3263,
13,
51376,
51524,
865,
11,
293,
550,
286,
2041,
754,
322,
264,
4829,
295,
411,
3825,
365,
3104,
281,
17443,
11,
51740,
51740
]
},
{
"avg_logprob": -0.26901139153374565,
"compression_ratio": 1.3588235294117648,
"end": 1319.52,
"id": 196,
"no_speech_prob": 0.00026941500254906714,
"seek": 130928,
"start": 1317.52,
"temperature": 0,
"text": " but that was explanatory.",
"tokens": [
50364,
1057,
558,
11,
3231,
13,
663,
390,
588,
4961,
13,
286,
632,
572,
1558,
577,
309,
1985,
294,
1729,
11,
50776,
50776,
457,
300,
390,
9045,
4745,
13,
50876,
51276,
27697,
286,
727,
652,
1101,
4337,
391,
3263,
13,
51376,
51524,
865,
11,
293,
550,
286,
2041,
754,
322,
264,
4829,
295,
411,
3825,
365,
3104,
281,
17443,
11,
51740,
51740
]
},
{
"avg_logprob": -0.26901139153374565,
"compression_ratio": 1.3588235294117648,
"end": 1329.52,
"id": 197,
"no_speech_prob": 0.00026941500254906714,
"seek": 130928,
"start": 1327.52,
"temperature": 0,
"text": " Wish I could make better otter sounds.",
"tokens": [
50364,
1057,
558,
11,
3231,
13,
663,
390,
588,
4961,
13,
286,
632,
572,
1558,
577,
309,
1985,
294,
1729,
11,
50776,
50776,
457,
300,
390,
9045,
4745,
13,
50876,
51276,
27697,
286,
727,
652,
1101,
4337,
391,
3263,
13,
51376,
51524,
865,
11,
293,
550,
286,
2041,
754,
322,
264,
4829,
295,
411,
3825,
365,
3104,
281,
17443,
11,
51740,
51740
]
},
{
"avg_logprob": -0.26901139153374565,
"compression_ratio": 1.3588235294117648,
"end": 1336.8,
"id": 198,
"no_speech_prob": 0.00026941500254906714,
"seek": 130928,
"start": 1332.48,
"temperature": 0,
"text": " Yeah, and then I guess even on the topic of like security with respect to weights,",
"tokens": [
50364,
1057,
558,
11,
3231,
13,
663,
390,
588,
4961,
13,
286,
632,
572,
1558,
577,
309,
1985,
294,
1729,
11,
50776,
50776,
457,
300,
390,
9045,
4745,
13,
50876,
51276,
27697,
286,
727,
652,
1101,
4337,
391,
3263,
13,
51376,
51524,
865,
11,
293,
550,
286,
2041,
754,
322,
264,
4829,
295,
411,
3825,
365,
3104,
281,
17443,
11,
51740,
51740
]
},
{
"avg_logprob": -0.1561054749922319,
"compression_ratio": 1.7214285714285715,
"end": 1341.68,
"id": 199,
"no_speech_prob": 0.00009760833199834451,
"seek": 133680,
"start": 1336.8,
"temperature": 0,
"text": " I mean, as Jason has been alluding to, I think multiple times now, like adversarial attacks are",
"tokens": [
50364,
286,
914,
11,
382,
11181,
575,
668,
439,
33703,
281,
11,
286,
519,
3866,
1413,
586,
11,
411,
17641,
44745,
8122,
366,
50608,
50608,
534,
11,
534,
1578,
294,
264,
2020,
300,
2318,
337,
3256,
17641,
44745,
8122,
420,
6278,
17641,
44745,
50884,
50884,
8122,
11,
291,
393,
40468,
264,
4846,
6358,
538,
364,
10100,
1336,
964,
2372,
281,
257,
1952,
293,
362,
264,
51180,
51180,
2316,
5598,
365,
411,
11803,
13,
23,
4,
6687,
11,
445,
257,
3879,
819,
21538,
1508,
11,
420,
362,
51488,
51488,
309,
1121,
735,
1936,
2035,
1230,
291,
528,
309,
281,
5598,
13,
400,
341,
307,
1936,
257,
18326,
295,
51812,
51812
]
},
{
"avg_logprob": -0.1561054749922319,
"compression_ratio": 1.7214285714285715,
"end": 1347.2,
"id": 200,
"no_speech_prob": 0.00009760833199834451,
"seek": 133680,
"start": 1341.68,
"temperature": 0,
"text": " really, really bad in the sense that especially for image adversarial attacks or audio adversarial",
"tokens": [
50364,
286,
914,
11,
382,
11181,
575,
668,
439,
33703,
281,
11,
286,
519,
3866,
1413,
586,
11,
411,
17641,
44745,
8122,
366,
50608,
50608,
534,
11,
534,
1578,
294,
264,
2020,
300,
2318,
337,
3256,
17641,
44745,
8122,
420,
6278,
17641,
44745,
50884,
50884,
8122,
11,
291,
393,
40468,
264,
4846,
6358,
538,
364,
10100,
1336,
964,
2372,
281,
257,
1952,
293,
362,
264,
51180,
51180,
2316,
5598,
365,
411,
11803,
13,
23,
4,
6687,
11,
445,
257,
3879,
819,
21538,
1508,
11,
420,
362,
51488,
51488,
309,
1121,
735,
1936,
2035,
1230,
291,
528,
309,
281,
5598,
13,
400,
341,
307,
1936,
257,
18326,
295,
51812,
51812
]
},
{
"avg_logprob": -0.1561054749922319,
"compression_ratio": 1.7214285714285715,
"end": 1353.12,
"id": 201,
"no_speech_prob": 0.00009760833199834451,
"seek": 133680,
"start": 1347.2,
"temperature": 0,
"text": " attacks, you can perturb the input signal by an imperceptible amount to a human and have the",
"tokens": [
50364,
286,
914,
11,
382,
11181,
575,
668,
439,
33703,
281,
11,
286,
519,
3866,
1413,
586,
11,
411,
17641,
44745,
8122,
366,
50608,
50608,
534,
11,
534,
1578,
294,
264,
2020,
300,
2318,
337,
3256,
17641,
44745,
8122,
420,
6278,
17641,
44745,
50884,
50884,
8122,
11,
291,
393,
40468,
264,
4846,
6358,
538,
364,
10100,
1336,
964,
2372,
281,
257,
1952,
293,
362,
264,
51180,
51180,
2316,
5598,
365,
411,
11803,
13,
23,
4,
6687,
11,
445,
257,
3879,
819,
21538,
1508,
11,
420,
362,
51488,
51488,
309,
1121,
735,
1936,
2035,
1230,
291,
528,
309,
281,
5598,
13,
400,
341,
307,
1936,
257,
18326,
295,
51812,
51812
]
},
{
"avg_logprob": -0.1561054749922319,
"compression_ratio": 1.7214285714285715,
"end": 1359.28,
"id": 202,
"no_speech_prob": 0.00009760833199834451,
"seek": 133680,
"start": 1353.12,
"temperature": 0,
"text": " model output with like 99.8% confidence, just a totally different classification class, or have",
"tokens": [
50364,
286,
914,
11,
382,
11181,
575,
668,
439,
33703,
281,
11,
286,
519,
3866,
1413,
586,
11,
411,
17641,
44745,
8122,
366,
50608,
50608,
534,
11,
534,
1578,
294,
264,
2020,
300,
2318,
337,
3256,
17641,
44745,
8122,
420,
6278,
17641,
44745,
50884,
50884,
8122,
11,
291,
393,
40468,
264,
4846,
6358,
538,
364,
10100,
1336,
964,
2372,
281,
257,
1952,
293,
362,
264,
51180,
51180,
2316,
5598,
365,
411,
11803,
13,
23,
4,
6687,
11,
445,
257,
3879,
819,
21538,
1508,
11,
420,
362,
51488,
51488,
309,
1121,
735,
1936,
2035,
1230,
291,
528,
309,
281,
5598,
13,
400,
341,
307,
1936,
257,
18326,
295,
51812,
51812
]
},
{
"avg_logprob": -0.1561054749922319,
"compression_ratio": 1.7214285714285715,
"end": 1365.76,
"id": 203,
"no_speech_prob": 0.00009760833199834451,
"seek": 133680,
"start": 1359.28,
"temperature": 0,
"text": " it regress basically whatever number you want it to output. And this is basically a consequence of",
"tokens": [
50364,
286,
914,
11,
382,
11181,
575,
668,
439,
33703,
281,
11,
286,
519,
3866,
1413,
586,
11,
411,
17641,
44745,
8122,
366,
50608,
50608,
534,
11,
534,
1578,
294,
264,
2020,
300,
2318,
337,
3256,
17641,
44745,
8122,
420,
6278,
17641,
44745,
50884,
50884,
8122,
11,
291,
393,
40468,
264,
4846,
6358,
538,
364,
10100,
1336,
964,
2372,
281,
257,
1952,
293,
362,
264,
51180,
51180,
2316,
5598,
365,
411,
11803,
13,
23,
4,
6687,
11,
445,
257,
3879,
819,
21538,
1508,
11,
420,
362,
51488,
51488,
309,
1121,
735,
1936,
2035,
1230,
291,
528,
309,
281,
5598,
13,
400,
341,
307,
1936,
257,
18326,
295,
51812,
51812
]
},
{
"avg_logprob": -0.14702268477973587,
"compression_ratio": 1.72992700729927,
"end": 1370.56,
"id": 204,
"no_speech_prob": 0.00017397876945324242,
"seek": 136576,
"start": 1365.76,
"temperature": 0,
"text": " the fact that models are differentiable and you can differentiate with respect to the outputs in",
"tokens": [
50364,
264,
1186,
300,
5245,
366,
819,
9364,
293,
291,
393,
23203,
365,
3104,
281,
264,
23930,
294,
50604,
50604,
257,
1629,
733,
295,
11,
321,
818,
309,
411,
26231,
16235,
23475,
636,
11,
281,
733,
295,
17522,
264,
50852,
50852,
2712,
11,
411,
264,
1952,
43276,
964,
2712,
322,
264,
4846,
1339,
920,
4473,
264,
5598,
1508,
13,
51120,
51160,
400,
370,
11,
1338,
11,
286,
2041,
264,
935,
307,
294,
341,
1389,
11,
498,
291,
458,
264,
2316,
17443,
420,
498,
291,
362,
257,
51428,
51428,
665,
28023,
295,
264,
2316,
17443,
11,
291,
393,
6018,
257,
2418,
2424,
2690,
11,
597,
1355,
16143,
51668,
51668
]
},
{
"avg_logprob": -0.14702268477973587,
"compression_ratio": 1.72992700729927,
"end": 1375.52,
"id": 205,
"no_speech_prob": 0.00017397876945324242,
"seek": 136576,
"start": 1370.56,
"temperature": 0,
"text": " a certain kind of, we call it like projected gradient descent way, to kind of minimize the",
"tokens": [
50364,
264,
1186,
300,
5245,
366,
819,
9364,
293,
291,
393,
23203,
365,
3104,
281,
264,
23930,
294,
50604,
50604,
257,
1629,
733,
295,
11,
321,
818,
309,
411,
26231,
16235,
23475,
636,
11,
281,
733,
295,
17522,
264,
50852,
50852,
2712,
11,
411,
264,
1952,
43276,
964,
2712,
322,
264,
4846,
1339,
920,
4473,
264,
5598,
1508,
13,
51120,
51160,
400,
370,
11,
1338,
11,
286,
2041,
264,
935,
307,
294,
341,
1389,
11,
498,
291,
458,
264,
2316,
17443,
420,
498,
291,
362,
257,
51428,
51428,
665,
28023,
295,
264,
2316,
17443,
11,
291,
393,
6018,
257,
2418,
2424,
2690,
11,
597,
1355,
16143,
51668,
51668
]
},
{
"avg_logprob": -0.14702268477973587,
"compression_ratio": 1.72992700729927,
"end": 1380.8799999999999,
"id": 206,
"no_speech_prob": 0.00017397876945324242,
"seek": 136576,
"start": 1375.52,
"temperature": 0,
"text": " impact, like the human perceptible impact on the input while still changing the output class.",
"tokens": [
50364,
264,
1186,
300,
5245,
366,
819,
9364,
293,
291,
393,
23203,
365,
3104,
281,
264,
23930,
294,
50604,
50604,
257,
1629,
733,
295,
11,
321,
818,
309,
411,
26231,
16235,
23475,
636,
11,
281,
733,
295,
17522,
264,
50852,
50852,
2712,
11,
411,
264,
1952,
43276,
964,
2712,
322,
264,
4846,
1339,
920,
4473,
264,
5598,
1508,
13,
51120,
51160,
400,
370,
11,
1338,
11,
286,
2041,
264,
935,
307,
294,
341,
1389,
11,
498,
291,
458,
264,
2316,
17443,
420,
498,
291,
362,
257,
51428,
51428,
665,
28023,
295,
264,
2316,
17443,
11,
291,
393,
6018,
257,
2418,
2424,
2690,
11,
597,
1355,
16143,
51668,
51668
]
},
{
"avg_logprob": -0.14702268477973587,
"compression_ratio": 1.72992700729927,
"end": 1387.04,
"id": 207,
"no_speech_prob": 0.00017397876945324242,
"seek": 136576,
"start": 1381.68,
"temperature": 0,
"text": " And so, yeah, I guess the point is in this case, if you know the model weights or if you have a",
"tokens": [
50364,
264,
1186,
300,
5245,
366,
819,
9364,
293,
291,
393,
23203,
365,
3104,
281,
264,
23930,
294,
50604,
50604,
257,
1629,
733,
295,
11,
321,
818,
309,
411,
26231,
16235,
23475,
636,
11,
281,
733,
295,
17522,
264,
50852,
50852,
2712,
11,
411,
264,
1952,
43276,
964,
2712,
322,
264,
4846,
1339,
920,
4473,
264,
5598,
1508,
13,
51120,
51160,
400,
370,
11,
1338,
11,
286,
2041,
264,
935,
307,
294,
341,
1389,
11,
498,
291,
458,
264,
2316,
17443,
420,
498,
291,
362,
257,
51428,
51428,
665,
28023,
295,
264,
2316,
17443,
11,
291,
393,
6018,
257,
2418,
2424,
2690,
11,
597,
1355,
16143,
51668,
51668
]
},
{
"avg_logprob": -0.14702268477973587,
"compression_ratio": 1.72992700729927,
"end": 1391.84,
"id": 208,
"no_speech_prob": 0.00017397876945324242,
"seek": 136576,
"start": 1387.04,
"temperature": 0,
"text": " good approximation of the model weights, you can conduct a white box attack, which means locally",
"tokens": [
50364,
264,
1186,
300,
5245,
366,
819,
9364,
293,
291,
393,
23203,
365,
3104,
281,
264,
23930,
294,
50604,
50604,
257,
1629,
733,
295,
11,
321,
818,
309,
411,
26231,
16235,
23475,
636,
11,
281,
733,
295,
17522,
264,
50852,
50852,
2712,
11,
411,
264,
1952,
43276,
964,
2712,
322,
264,
4846,
1339,
920,
4473,
264,
5598,
1508,
13,
51120,
51160,
400,
370,
11,
1338,
11,
286,
2041,
264,
935,
307,
294,
341,
1389,
11,
498,
291,
458,
264,
2316,
17443,
420,
498,
291,
362,
257,
51428,
51428,
665,
28023,
295,
264,
2316,
17443,
11,
291,
393,
6018,
257,
2418,
2424,
2690,
11,
597,
1355,
16143,
51668,
51668
]
},
{
"avg_logprob": -0.1514128544291512,
"compression_ratio": 1.685512367491166,
"end": 1397.36,
"id": 209,
"no_speech_prob": 0.000012029498066112865,
"seek": 139184,
"start": 1391.84,
"temperature": 0,
"text": " you just run the attack and then you are guaranteed basically that the model will accept it.",
"tokens": [
50364,
291,
445,
1190,
264,
2690,
293,
550,
291,
366,
18031,
1936,
300,
264,
2316,
486,
3241,
309,
13,
50640,
50640,
400,
309,
311,
472,
295,
264,
721,
689,
309,
311,
733,
295,
257,
3834,
1257,
3004,
10576,
570,
1176,
42,
12683,
393,
3191,
341,
50968,
50968,
1154,
538,
8295,
291,
6479,
264,
17443,
13,
467,
311,
406,
411,
257,
2176,
11632,
1970,
17641,
44745,
8122,
11,
51180,
51180,
457,
309,
3665,
257,
688,
13,
583,
550,
291,
3624,
484,
322,
733,
295,
264,
3670,
1737,
6468,
329,
295,
2878,
885,
1075,
281,
1190,
51464,
51464,
264,
2316,
13,
407,
309,
311,
11,
291,
393,
909,
264,
17443,
11,
457,
550,
291,
643,
281,
5777,
1125,
264,
12973,
11,
51692,
51692
]
},
{
"avg_logprob": -0.1514128544291512,
"compression_ratio": 1.685512367491166,
"end": 1403.9199999999998,
"id": 210,
"no_speech_prob": 0.000012029498066112865,
"seek": 139184,
"start": 1397.36,
"temperature": 0,
"text": " And it's one of the things where it's kind of a double edged sword because ZKML can fix this",
"tokens": [
50364,
291,
445,
1190,
264,
2690,
293,
550,
291,
366,
18031,
1936,
300,
264,
2316,
486,
3241,
309,
13,
50640,
50640,
400,
309,
311,
472,
295,
264,
721,
689,
309,
311,
733,
295,
257,
3834,
1257,
3004,
10576,
570,
1176,
42,
12683,
393,
3191,
341,
50968,
50968,
1154,
538,
8295,
291,
6479,
264,
17443,
13,
467,
311,
406,
411,
257,
2176,
11632,
1970,
17641,
44745,
8122,
11,
51180,
51180,
457,
309,
3665,
257,
688,
13,
583,
550,
291,
3624,
484,
322,
733,
295,
264,
3670,
1737,
6468,
329,
295,
2878,
885,
1075,
281,
1190,
51464,
51464,
264,
2316,
13,
407,
309,
311,
11,
291,
393,
909,
264,
17443,
11,
457,
550,
291,
643,
281,
5777,
1125,
264,
12973,
11,
51692,
51692
]
},
{
"avg_logprob": -0.1514128544291512,
"compression_ratio": 1.685512367491166,
"end": 1408.1599999999999,
"id": 211,
"no_speech_prob": 0.000012029498066112865,
"seek": 139184,
"start": 1403.9199999999998,
"temperature": 0,
"text": " problem by letting you hide the weights. It's not like a perfect bullet against adversarial attacks,",
"tokens": [
50364,
291,
445,
1190,
264,
2690,
293,
550,
291,
366,
18031,
1936,
300,
264,
2316,
486,
3241,
309,
13,
50640,
50640,
400,
309,
311,
472,
295,
264,
721,
689,
309,
311,
733,
295,
257,
3834,
1257,
3004,
10576,
570,
1176,
42,
12683,
393,
3191,
341,
50968,
50968,
1154,
538,
8295,
291,
6479,
264,
17443,
13,
467,
311,
406,
411,
257,
2176,
11632,
1970,
17641,
44745,
8122,
11,
51180,
51180,
457,
309,
3665,
257,
688,
13,
583,
550,
291,
3624,
484,
322,
733,
295,
264,
3670,
1737,
6468,
329,
295,
2878,
885,
1075,
281,
1190,
51464,
51464,
264,
2316,
13,
407,
309,
311,
11,
291,
393,
909,
264,
17443,
11,
457,
550,
291,
643,
281,
5777,
1125,
264,
12973,
11,
51692,
51692
]
},
{
"avg_logprob": -0.1514128544291512,
"compression_ratio": 1.685512367491166,
"end": 1413.84,
"id": 212,
"no_speech_prob": 0.000012029498066112865,
"seek": 139184,
"start": 1408.1599999999999,
"temperature": 0,
"text": " but it helps a lot. But then you lose out on kind of the web free ethos of anyone being able to run",
"tokens": [
50364,
291,
445,
1190,
264,
2690,
293,
550,
291,
366,
18031,
1936,
300,
264,
2316,
486,
3241,
309,
13,
50640,
50640,
400,
309,
311,
472,
295,
264,
721,
689,
309,
311,
733,
295,
257,
3834,
1257,
3004,
10576,
570,
1176,
42,
12683,
393,
3191,
341,
50968,
50968,
1154,
538,
8295,
291,
6479,
264,
17443,
13,
467,
311,
406,
411,
257,
2176,
11632,
1970,
17641,
44745,
8122,
11,
51180,
51180,
457,
309,
3665,
257,
688,
13,
583,
550,
291,
3624,
484,
322,
733,
295,
264,
3670,
1737,
6468,
329,
295,
2878,
885,
1075,
281,
1190,
51464,
51464,
264,
2316,
13,
407,
309,
311,
11,
291,
393,
909,
264,
17443,
11,
457,
550,
291,
643,
281,
5777,
1125,
264,
12973,
11,
51692,
51692
]
},
{
"avg_logprob": -0.1514128544291512,
"compression_ratio": 1.685512367491166,
"end": 1418.3999999999999,
"id": 213,
"no_speech_prob": 0.000012029498066112865,
"seek": 139184,
"start": 1413.84,
"temperature": 0,
"text": " the model. So it's, you can add the weights, but then you need to centralize the operator,",
"tokens": [
50364,
291,
445,
1190,
264,
2690,
293,
550,
291,
366,
18031,
1936,
300,
264,
2316,
486,
3241,
309,
13,
50640,
50640,
400,
309,
311,
472,
295,
264,
721,
689,
309,
311,
733,
295,
257,
3834,
1257,
3004,
10576,
570,
1176,
42,
12683,
393,
3191,
341,
50968,
50968,
1154,
538,
8295,
291,
6479,
264,
17443,
13,
467,
311,
406,
411,
257,
2176,
11632,
1970,
17641,
44745,
8122,
11,
51180,
51180,
457,
309,
3665,
257,
688,
13,
583,
550,
291,
3624,
484,
322,
733,
295,
264,
3670,
1737,
6468,
329,
295,
2878,
885,
1075,
281,
1190,
51464,
51464,
264,
2316,
13,
407,
309,
311,
11,
291,
393,
909,
264,
17443,
11,
457,
550,
291,
643,
281,
5777,
1125,
264,
12973,
11,
51692,
51692
]
},
{
"avg_logprob": -0.22153906236615098,
"compression_ratio": 1.6569343065693432,
"end": 1422,
"id": 214,
"no_speech_prob": 0.00007030949200270697,
"seek": 141840,
"start": 1418.4,
"temperature": 0,
"text": " or you cannot hide the weights, but then you have to deal with adversarial effects.",
"tokens": [
50364,
420,
291,
2644,
6479,
264,
17443,
11,
457,
550,
291,
362,
281,
2028,
365,
17641,
44745,
5065,
13,
50544,
50544,
407,
411,
472,
1194,
294,
341,
3513,
1062,
754,
281,
11,
286,
2041,
11,
362,
411,
2107,
12,
49136,
259,
3142,
5245,
420,
50924,
50924,
1310,
5245,
300,
3212,
380,
370,
3612,
646,
447,
3320,
807,
11,
558,
30,
407,
411,
460,
1770,
82,
11,
337,
1365,
11,
362,
51224,
51224,
257,
21179,
1823,
11,
597,
291,
393,
11,
286,
2041,
11,
30874,
294,
512,
2020,
365,
411,
512,
5508,
2445,
13,
51464,
51500,
583,
294,
2674,
11,
498,
321,
411,
829,
1547,
31420,
84,
1088,
420,
3847,
5245,
689,
51756,
51784
]
},
{
"avg_logprob": -0.22153906236615098,
"compression_ratio": 1.6569343065693432,
"end": 1429.6000000000001,
"id": 215,
"no_speech_prob": 0.00007030949200270697,
"seek": 141840,
"start": 1422,
"temperature": 0,
"text": " So like one thought in this direction might even to, I guess, have like non-deterministic models or",
"tokens": [
50364,
420,
291,
2644,
6479,
264,
17443,
11,
457,
550,
291,
362,
281,
2028,
365,
17641,
44745,
5065,
13,
50544,
50544,
407,
411,
472,
1194,
294,
341,
3513,
1062,
754,
281,
11,
286,
2041,
11,
362,
411,
2107,
12,
49136,
259,
3142,
5245,
420,
50924,
50924,
1310,
5245,
300,
3212,
380,
370,
3612,
646,
447,
3320,
807,
11,
558,
30,
407,
411,
460,
1770,
82,
11,
337,
1365,
11,
362,
51224,
51224,
257,
21179,
1823,
11,
597,
291,
393,
11,
286,
2041,
11,
30874,
294,
512,
2020,
365,
411,
512,
5508,
2445,
13,
51464,
51500,
583,
294,
2674,
11,
498,
321,
411,
829,
1547,
31420,
84,
1088,
420,
3847,
5245,
689,
51756,
51784
]
},
{
"avg_logprob": -0.22153906236615098,
"compression_ratio": 1.6569343065693432,
"end": 1435.6000000000001,
"id": 216,
"no_speech_prob": 0.00007030949200270697,
"seek": 141840,
"start": 1429.6000000000001,
"temperature": 0,
"text": " maybe models that aren't so easily back propped through, right? So like GANs, for example, have",
"tokens": [
50364,
420,
291,
2644,
6479,
264,
17443,
11,
457,
550,
291,
362,
281,
2028,
365,
17641,
44745,
5065,
13,
50544,
50544,
407,
411,
472,
1194,
294,
341,
3513,
1062,
754,
281,
11,
286,
2041,
11,
362,
411,
2107,
12,
49136,
259,
3142,
5245,
420,
50924,
50924,
1310,
5245,
300,
3212,
380,
370,
3612,
646,
447,
3320,
807,
11,
558,
30,
407,
411,
460,
1770,
82,
11,
337,
1365,
11,
362,
51224,
51224,
257,
21179,
1823,
11,
597,
291,
393,
11,
286,
2041,
11,
30874,
294,
512,
2020,
365,
411,
512,
5508,
2445,
13,
51464,
51500,
583,
294,
2674,
11,
498,
321,
411,
829,
1547,
31420,
84,
1088,
420,
3847,
5245,
689,
51756,
51784
]
},
{
"avg_logprob": -0.22153906236615098,
"compression_ratio": 1.6569343065693432,
"end": 1440.4,
"id": 217,
"no_speech_prob": 0.00007030949200270697,
"seek": 141840,
"start": 1435.6000000000001,
"temperature": 0,
"text": " a sampling step, which you can, I guess, approximate in some sense with like some smooth function.",
"tokens": [
50364,
420,
291,
2644,
6479,
264,
17443,
11,
457,
550,
291,
362,
281,
2028,
365,
17641,
44745,
5065,
13,
50544,
50544,
407,
411,
472,
1194,
294,
341,
3513,
1062,
754,
281,
11,
286,
2041,
11,
362,
411,
2107,
12,
49136,
259,
3142,
5245,
420,
50924,
50924,
1310,
5245,
300,
3212,
380,
370,
3612,
646,
447,
3320,
807,
11,
558,
30,
407,
411,
460,
1770,
82,
11,
337,
1365,
11,
362,
51224,
51224,
257,
21179,
1823,
11,
597,
291,
393,
11,
286,
2041,
11,
30874,
294,
512,
2020,
365,
411,
512,
5508,
2445,
13,
51464,
51500,
583,
294,
2674,
11,
498,
321,
411,
829,
1547,
31420,
84,
1088,
420,
3847,
5245,
689,
51756,
51784
]
},
{
"avg_logprob": -0.22153906236615098,
"compression_ratio": 1.6569343065693432,
"end": 1446.24,
"id": 218,
"no_speech_prob": 0.00007030949200270697,
"seek": 141840,
"start": 1441.1200000000001,
"temperature": 0,
"text": " But in general, if we like put enough discontinuities or train models where",
"tokens": [
50364,
420,
291,
2644,
6479,
264,
17443,
11,
457,
550,
291,
362,
281,
2028,
365,
17641,
44745,
5065,
13,
50544,
50544,
407,
411,
472,
1194,
294,
341,
3513,
1062,
754,
281,
11,
286,
2041,
11,
362,
411,
2107,
12,
49136,
259,
3142,
5245,
420,
50924,
50924,
1310,
5245,
300,
3212,
380,
370,
3612,
646,
447,
3320,
807,
11,
558,
30,
407,
411,
460,
1770,
82,
11,
337,
1365,
11,
362,
51224,
51224,
257,
21179,
1823,
11,
597,
291,
393,
11,
286,
2041,
11,
30874,
294,
512,
2020,
365,
411,
512,
5508,
2445,
13,
51464,
51500,
583,
294,
2674,
11,
498,
321,
411,
829,
1547,
31420,
84,
1088,
420,
3847,
5245,
689,
51756,
51784
]
},
{
"avg_logprob": -0.19931476060734238,
"compression_ratio": 1.6563876651982379,
"end": 1451.92,
"id": 219,
"no_speech_prob": 0.000043317464587744325,
"seek": 144624,
"start": 1446.24,
"temperature": 0,
"text": " it's not so straightforward, I guess, like to go from output back to, or to go from changes in input",
"tokens": [
50364,
309,
311,
406,
370,
15325,
11,
286,
2041,
11,
411,
281,
352,
490,
5598,
646,
281,
11,
420,
281,
352,
490,
2962,
294,
4846,
50648,
50648,
420,
2962,
294,
19376,
281,
5598,
11,
550,
4099,
264,
17443,
1062,
767,
312,
1392,
13,
50892,
50892,
2014,
456,
7512,
281,
652,
264,
2418,
13058,
3366,
544,
28328,
281,
689,
411,
264,
46249,
307,
51348,
51348,
9212,
420,
689,
309,
311,
1333,
295,
257,
2684,
3779,
689,
309,
576,
11,
291,
576,
362,
281,
22277,
51660,
51712
]
},
{
"avg_logprob": -0.19931476060734238,
"compression_ratio": 1.6563876651982379,
"end": 1456.8,
"id": 220,
"no_speech_prob": 0.000043317464587744325,
"seek": 144624,
"start": 1451.92,
"temperature": 0,
"text": " or changes in intermediate to output, then showing the weights might actually be okay.",
"tokens": [
50364,
309,
311,
406,
370,
15325,
11,
286,
2041,
11,
411,
281,
352,
490,
5598,
646,
281,
11,
420,
281,
352,
490,
2962,
294,
4846,
50648,
50648,
420,
2962,
294,
19376,
281,
5598,
11,
550,
4099,
264,
17443,
1062,
767,
312,
1392,
13,
50892,
50892,
2014,
456,
7512,
281,
652,
264,
2418,
13058,
3366,
544,
28328,
281,
689,
411,
264,
46249,
307,
51348,
51348,
9212,
420,
689,
309,
311,
1333,
295,
257,
2684,
3779,
689,
309,
576,
11,
291,
576,
362,
281,
22277,
51660,
51712
]
},
{
"avg_logprob": -0.19931476060734238,
"compression_ratio": 1.6563876651982379,
"end": 1465.92,
"id": 221,
"no_speech_prob": 0.000043317464587744325,
"seek": 144624,
"start": 1456.8,
"temperature": 0,
"text": " Are there techniques to make the white bucket tax more costly to where like the profitability is",
"tokens": [
50364,
309,
311,
406,
370,
15325,
11,
286,
2041,
11,
411,
281,
352,
490,
5598,
646,
281,
11,
420,
281,
352,
490,
2962,
294,
4846,
50648,
50648,
420,
2962,
294,
19376,
281,
5598,
11,
550,
4099,
264,
17443,
1062,
767,
312,
1392,
13,
50892,
50892,
2014,
456,
7512,
281,
652,
264,
2418,
13058,
3366,
544,
28328,
281,
689,
411,
264,
46249,
307,
51348,
51348,
9212,
420,
689,
309,
311,
1333,
295,
257,
2684,
3779,
689,
309,
576,
11,
291,
576,
362,
281,
22277,
51660,
51712
]
},
{
"avg_logprob": -0.19931476060734238,
"compression_ratio": 1.6563876651982379,
"end": 1472.16,
"id": 222,
"no_speech_prob": 0.000043317464587744325,
"seek": 144624,
"start": 1465.92,
"temperature": 0,
"text": " reduced or where it's sort of a moving target where it would, you would have to continually",
"tokens": [
50364,
309,
311,
406,
370,
15325,
11,
286,
2041,
11,
411,
281,
352,
490,
5598,
646,
281,
11,
420,
281,
352,
490,
2962,
294,
4846,
50648,
50648,
420,
2962,
294,
19376,
281,
5598,
11,
550,
4099,
264,
17443,
1062,
767,
312,
1392,
13,
50892,
50892,
2014,
456,
7512,
281,
652,
264,
2418,
13058,
3366,
544,
28328,
281,
689,
411,
264,
46249,
307,
51348,
51348,
9212,
420,
689,
309,
311,
1333,
295,
257,
2684,
3779,
689,
309,
576,
11,
291,
576,
362,
281,
22277,
51660,
51712
]
},
{
"avg_logprob": -0.18667067014254057,
"compression_ratio": 1.6742424242424243,
"end": 1475.2,
"id": 223,
"no_speech_prob": 0.000007766869202896487,
"seek": 147216,
"start": 1472.16,
"temperature": 0,
"text": " recompute a new attack vector or a new attack strategy?",
"tokens": [
50364,
48000,
1169,
257,
777,
2690,
8062,
420,
257,
777,
2690,
5206,
30,
50516,
50604,
865,
11,
370,
294,
2674,
11,
15866,
364,
12692,
4846,
365,
512,
295,
729,
1785,
12,
2670,
12,
3322,
12,
446,
8122,
11,
370,
50956,
50956,
2361,
322,
437,
286,
1604,
11,
412,
1935,
456,
311,
341,
534,
11,
534,
2068,
2690,
1219,
14256,
3812,
12,
54,
22720,
51160,
51160,
2516,
411,
257,
1916,
4714,
1936,
411,
16235,
23475,
36540,
13,
407,
15866,
264,
51396,
51396,
2690,
307,
1238,
28328,
11,
293,
291,
393,
652,
309,
544,
28328,
538,
1419,
411,
257,
9521,
19492,
295,
5245,
51700,
51700
]
},
{
"avg_logprob": -0.18667067014254057,
"compression_ratio": 1.6742424242424243,
"end": 1484,
"id": 224,
"no_speech_prob": 0.000007766869202896487,
"seek": 147216,
"start": 1476.96,
"temperature": 0,
"text": " Yeah, so in general, computing an attacked input with some of those state-of-the-art attacks, so",
"tokens": [
50364,
48000,
1169,
257,
777,
2690,
8062,
420,
257,
777,
2690,
5206,
30,
50516,
50604,
865,
11,
370,
294,
2674,
11,
15866,
364,
12692,
4846,
365,
512,
295,
729,
1785,
12,
2670,
12,
3322,
12,
446,
8122,
11,
370,
50956,
50956,
2361,
322,
437,
286,
1604,
11,
412,
1935,
456,
311,
341,
534,
11,
534,
2068,
2690,
1219,
14256,
3812,
12,
54,
22720,
51160,
51160,
2516,
411,
257,
1916,
4714,
1936,
411,
16235,
23475,
36540,
13,
407,
15866,
264,
51396,
51396,
2690,
307,
1238,
28328,
11,
293,
291,
393,
652,
309,
544,
28328,
538,
1419,
411,
257,
9521,
19492,
295,
5245,
51700,
51700
]
},
{
"avg_logprob": -0.18667067014254057,
"compression_ratio": 1.6742424242424243,
"end": 1488.0800000000002,
"id": 225,
"no_speech_prob": 0.000007766869202896487,
"seek": 147216,
"start": 1484,
"temperature": 0,
"text": " based on what I remember, at least there's this really, really strong attack called Carlini-Wagner",
"tokens": [
50364,
48000,
1169,
257,
777,
2690,
8062,
420,
257,
777,
2690,
5206,
30,
50516,
50604,
865,
11,
370,
294,
2674,
11,
15866,
364,
12692,
4846,
365,
512,
295,
729,
1785,
12,
2670,
12,
3322,
12,
446,
8122,
11,
370,
50956,
50956,
2361,
322,
437,
286,
1604,
11,
412,
1935,
456,
311,
341,
534,
11,
534,
2068,
2690,
1219,
14256,
3812,
12,
54,
22720,
51160,
51160,
2516,
411,
257,
1916,
4714,
1936,
411,
16235,
23475,
36540,
13,
407,
15866,
264,
51396,
51396,
2690,
307,
1238,
28328,
11,
293,
291,
393,
652,
309,
544,
28328,
538,
1419,
411,
257,
9521,
19492,
295,
5245,
51700,
51700
]
},
{
"avg_logprob": -0.18667067014254057,
"compression_ratio": 1.6742424242424243,
"end": 1492.8000000000002,
"id": 226,
"no_speech_prob": 0.000007766869202896487,
"seek": 147216,
"start": 1488.0800000000002,
"temperature": 0,
"text": " takes like a couple thousand basically like gradient descent iterations. So computing the",
"tokens": [
50364,
48000,
1169,
257,
777,
2690,
8062,
420,
257,
777,
2690,
5206,
30,
50516,
50604,
865,
11,
370,
294,
2674,
11,
15866,
364,
12692,
4846,
365,
512,
295,
729,
1785,
12,
2670,
12,
3322,
12,
446,
8122,
11,
370,
50956,
50956,
2361,
322,
437,
286,
1604,
11,
412,
1935,
456,
311,
341,
534,
11,
534,
2068,
2690,
1219,
14256,
3812,
12,
54,
22720,
51160,
51160,
2516,
411,
257,
1916,
4714,
1936,
411,
16235,
23475,
36540,
13,
407,
15866,
264,
51396,
51396,
2690,
307,
1238,
28328,
11,
293,
291,
393,
652,
309,
544,
28328,
538,
1419,
411,
257,
9521,
19492,
295,
5245,
51700,
51700
]
},
{
"avg_logprob": -0.18667067014254057,
"compression_ratio": 1.6742424242424243,
"end": 1498.88,
"id": 227,
"no_speech_prob": 0.000007766869202896487,
"seek": 147216,
"start": 1492.8000000000002,
"temperature": 0,
"text": " attack is pretty costly, and you can make it more costly by having like a diverse ensemble of models",
"tokens": [
50364,
48000,
1169,
257,
777,
2690,
8062,
420,
257,
777,
2690,
5206,
30,
50516,
50604,
865,
11,
370,
294,
2674,
11,
15866,
364,
12692,
4846,
365,
512,
295,
729,
1785,
12,
2670,
12,
3322,
12,
446,
8122,
11,
370,
50956,
50956,
2361,
322,
437,
286,
1604,
11,
412,
1935,
456,
311,
341,
534,
11,
534,
2068,
2690,
1219,
14256,
3812,
12,
54,
22720,
51160,
51160,
2516,
411,
257,
1916,
4714,
1936,
411,
16235,
23475,
36540,
13,
407,
15866,
264,
51396,
51396,
2690,
307,
1238,
28328,
11,
293,
291,
393,
652,
309,
544,
28328,
538,
1419,
411,
257,
9521,
19492,
295,
5245,
51700,
51700
]
},
{
"avg_logprob": -0.18977737426757812,
"compression_ratio": 1.7686567164179106,
"end": 1503.68,
"id": 228,
"no_speech_prob": 0.000025070794436032884,
"seek": 149888,
"start": 1498.88,
"temperature": 0,
"text": " for example, where an attack which works very well for one of the models doesn't work very well for",
"tokens": [
50364,
337,
1365,
11,
689,
364,
2690,
597,
1985,
588,
731,
337,
472,
295,
264,
5245,
1177,
380,
589,
588,
731,
337,
50604,
50604,
264,
2357,
570,
291,
366,
787,
16235,
40182,
365,
3104,
281,
472,
295,
264,
5245,
13,
50808,
50832,
2022,
300,
885,
848,
11,
2211,
2424,
8122,
366,
611,
7970,
534,
665,
11,
597,
307,
1687,
11,
51096,
51096,
1687,
4227,
13,
440,
1558,
885,
300,
291,
393,
445,
3847,
257,
2531,
2316,
322,
8659,
264,
912,
1412,
992,
51396,
51396,
300,
264,
2316,
291,
434,
15010,
307,
11,
390,
8895,
322,
293,
483,
746,
411,
787,
411,
257,
1922,
12,
986,
68,
15485,
51696,
51696
]
},
{
"avg_logprob": -0.18977737426757812,
"compression_ratio": 1.7686567164179106,
"end": 1507.7600000000002,
"id": 229,
"no_speech_prob": 0.000025070794436032884,
"seek": 149888,
"start": 1503.68,
"temperature": 0,
"text": " the others because you are only gradient descending with respect to one of the models.",
"tokens": [
50364,
337,
1365,
11,
689,
364,
2690,
597,
1985,
588,
731,
337,
472,
295,
264,
5245,
1177,
380,
589,
588,
731,
337,
50604,
50604,
264,
2357,
570,
291,
366,
787,
16235,
40182,
365,
3104,
281,
472,
295,
264,
5245,
13,
50808,
50832,
2022,
300,
885,
848,
11,
2211,
2424,
8122,
366,
611,
7970,
534,
665,
11,
597,
307,
1687,
11,
51096,
51096,
1687,
4227,
13,
440,
1558,
885,
300,
291,
393,
445,
3847,
257,
2531,
2316,
322,
8659,
264,
912,
1412,
992,
51396,
51396,
300,
264,
2316,
291,
434,
15010,
307,
11,
390,
8895,
322,
293,
483,
746,
411,
787,
411,
257,
1922,
12,
986,
68,
15485,
51696,
51696
]
},
{
"avg_logprob": -0.18977737426757812,
"compression_ratio": 1.7686567164179106,
"end": 1513.5200000000002,
"id": 230,
"no_speech_prob": 0.000025070794436032884,
"seek": 149888,
"start": 1508.24,
"temperature": 0,
"text": " With that being said, black box attacks are also apparently really good, which is super,",
"tokens": [
50364,
337,
1365,
11,
689,
364,
2690,
597,
1985,
588,
731,
337,
472,
295,
264,
5245,
1177,
380,
589,
588,
731,
337,
50604,
50604,
264,
2357,
570,
291,
366,
787,
16235,
40182,
365,
3104,
281,
472,
295,
264,
5245,
13,
50808,
50832,
2022,
300,
885,
848,
11,
2211,
2424,
8122,
366,
611,
7970,
534,
665,
11,
597,
307,
1687,
11,
51096,
51096,
1687,
4227,
13,
440,
1558,
885,
300,
291,
393,
445,
3847,
257,
2531,
2316,
322,
8659,
264,
912,
1412,
992,
51396,
51396,
300,
264,
2316,
291,
434,
15010,
307,
11,
390,
8895,
322,
293,
483,
746,
411,
787,
411,
257,
1922,
12,
986,
68,
15485,
51696,
51696
]
},
{
"avg_logprob": -0.18977737426757812,
"compression_ratio": 1.7686567164179106,
"end": 1519.5200000000002,
"id": 231,
"no_speech_prob": 0.000025070794436032884,
"seek": 149888,
"start": 1513.5200000000002,
"temperature": 0,
"text": " super sad. The idea being that you can just train a similar model on effectively the same data set",
"tokens": [
50364,
337,
1365,
11,
689,
364,
2690,
597,
1985,
588,
731,
337,
472,
295,
264,
5245,
1177,
380,
589,
588,
731,
337,
50604,
50604,
264,
2357,
570,
291,
366,
787,
16235,
40182,
365,
3104,
281,
472,
295,
264,
5245,
13,
50808,
50832,
2022,
300,
885,
848,
11,
2211,
2424,
8122,
366,
611,
7970,
534,
665,
11,
597,
307,
1687,
11,
51096,
51096,
1687,
4227,
13,
440,
1558,
885,
300,
291,
393,
445,
3847,
257,
2531,
2316,
322,
8659,
264,
912,
1412,
992,
51396,
51396,
300,
264,
2316,
291,
434,
15010,
307,
11,
390,
8895,
322,
293,
483,
746,
411,
787,
411,
257,
1922,
12,
986,
68,
15485,
51696,
51696
]
},
{
"avg_logprob": -0.18977737426757812,
"compression_ratio": 1.7686567164179106,
"end": 1525.5200000000002,
"id": 232,
"no_speech_prob": 0.000025070794436032884,
"seek": 149888,
"start": 1519.5200000000002,
"temperature": 0,
"text": " that the model you're attacking is, was trained on and get something like only like a half-redeemed",
"tokens": [
50364,
337,
1365,
11,
689,
364,
2690,
597,
1985,
588,
731,
337,
472,
295,
264,
5245,
1177,
380,
589,
588,
731,
337,
50604,
50604,
264,
2357,
570,
291,
366,
787,
16235,
40182,
365,
3104,
281,
472,
295,
264,
5245,
13,
50808,
50832,
2022,
300,
885,
848,
11,
2211,
2424,
8122,
366,
611,
7970,
534,
665,
11,
597,
307,
1687,
11,
51096,
51096,
1687,
4227,
13,
440,
1558,
885,
300,
291,
393,
445,
3847,
257,
2531,
2316,
322,
8659,
264,
912,
1412,
992,
51396,
51396,
300,
264,
2316,
291,
434,
15010,
307,
11,
390,
8895,
322,
293,
483,
746,
411,
787,
411,
257,
1922,
12,
986,
68,
15485,
51696,
51696
]
},
{
"avg_logprob": -0.10461953255982526,
"compression_ratio": 1.746212121212121,
"end": 1530.4,
"id": 233,
"no_speech_prob": 0.000015445955796167254,
"seek": 152552,
"start": 1525.52,
"temperature": 0,
"text": " was trained on and get something like only like a half reduction in your attack success.",
"tokens": [
50364,
390,
8895,
322,
293,
483,
746,
411,
787,
411,
257,
1922,
11004,
294,
428,
2690,
2245,
13,
50608,
50644,
467,
311,
445,
4930,
11,
309,
311,
1596,
4930,
13,
286,
2041,
8328,
952,
311,
935,
295,
577,
867,
295,
613,
50924,
50924,
2690,
712,
5245,
293,
577,
867,
295,
613,
8122,
366,
1944,
294,
4265,
1496,
5245,
11,
689,
411,
51224,
51224,
264,
2316,
2564,
307,
7633,
11,
291,
500,
380,
534,
458,
577,
411,
264,
1412,
307,
16146,
11,
498,
264,
1412,
307,
51444,
51444,
48704,
11,
721,
411,
300,
307,
4961,
11,
558,
30,
11164,
538,
22082,
3051,
13,
583,
412,
264,
917,
295,
264,
786,
11,
51772,
51812
]
},
{
"avg_logprob": -0.10461953255982526,
"compression_ratio": 1.746212121212121,
"end": 1536.72,
"id": 234,
"no_speech_prob": 0.000015445955796167254,
"seek": 152552,
"start": 1531.12,
"temperature": 0,
"text": " It's just tough, it's quite tough. I guess Florian's point of how many of these",
"tokens": [
50364,
390,
8895,
322,
293,
483,
746,
411,
787,
411,
257,
1922,
11004,
294,
428,
2690,
2245,
13,
50608,
50644,
467,
311,
445,
4930,
11,
309,
311,
1596,
4930,
13,
286,
2041,
8328,
952,
311,
935,
295,
577,
867,
295,
613,
50924,
50924,
2690,
712,
5245,
293,
577,
867,
295,
613,
8122,
366,
1944,
294,
4265,
1496,
5245,
11,
689,
411,
51224,
51224,
264,
2316,
2564,
307,
7633,
11,
291,
500,
380,
534,
458,
577,
411,
264,
1412,
307,
16146,
11,
498,
264,
1412,
307,
51444,
51444,
48704,
11,
721,
411,
300,
307,
4961,
11,
558,
30,
11164,
538,
22082,
3051,
13,
583,
412,
264,
917,
295,
264,
786,
11,
51772,
51812
]
},
{
"avg_logprob": -0.10461953255982526,
"compression_ratio": 1.746212121212121,
"end": 1542.72,
"id": 235,
"no_speech_prob": 0.000015445955796167254,
"seek": 152552,
"start": 1536.72,
"temperature": 0,
"text": " attackable models and how many of these attacks are possible in production level models, where like",
"tokens": [
50364,
390,
8895,
322,
293,
483,
746,
411,
787,
411,
257,
1922,
11004,
294,
428,
2690,
2245,
13,
50608,
50644,
467,
311,
445,
4930,
11,
309,
311,
1596,
4930,
13,
286,
2041,
8328,
952,
311,
935,
295,
577,
867,
295,
613,
50924,
50924,
2690,
712,
5245,
293,
577,
867,
295,
613,
8122,
366,
1944,
294,
4265,
1496,
5245,
11,
689,
411,
51224,
51224,
264,
2316,
2564,
307,
7633,
11,
291,
500,
380,
534,
458,
577,
411,
264,
1412,
307,
16146,
11,
498,
264,
1412,
307,
51444,
51444,
48704,
11,
721,
411,
300,
307,
4961,
11,
558,
30,
11164,
538,
22082,
3051,
13,
583,
412,
264,
917,
295,
264,
786,
11,
51772,
51812
]
},
{
"avg_logprob": -0.10461953255982526,
"compression_ratio": 1.746212121212121,
"end": 1547.12,
"id": 236,
"no_speech_prob": 0.000015445955796167254,
"seek": 152552,
"start": 1542.72,
"temperature": 0,
"text": " the model itself is hidden, you don't really know how like the data is cleaned, if the data is",
"tokens": [
50364,
390,
8895,
322,
293,
483,
746,
411,
787,
411,
257,
1922,
11004,
294,
428,
2690,
2245,
13,
50608,
50644,
467,
311,
445,
4930,
11,
309,
311,
1596,
4930,
13,
286,
2041,
8328,
952,
311,
935,
295,
577,
867,
295,
613,
50924,
50924,
2690,
712,
5245,
293,
577,
867,
295,
613,
8122,
366,
1944,
294,
4265,
1496,
5245,
11,
689,
411,
51224,
51224,
264,
2316,
2564,
307,
7633,
11,
291,
500,
380,
534,
458,
577,
411,
264,
1412,
307,
16146,
11,
498,
264,
1412,
307,
51444,
51444,
48704,
11,
721,
411,
300,
307,
4961,
11,
558,
30,
11164,
538,
22082,
3051,
13,
583,
412,
264,
917,
295,
264,
786,
11,
51772,
51812
]
},
{
"avg_logprob": -0.10461953255982526,
"compression_ratio": 1.746212121212121,
"end": 1553.68,
"id": 237,
"no_speech_prob": 0.000015445955796167254,
"seek": 152552,
"start": 1547.12,
"temperature": 0,
"text": " normalized, things like that is helpful, right? Security by obscurity. But at the end of the day,",
"tokens": [
50364,
390,
8895,
322,
293,
483,
746,
411,
787,
411,
257,
1922,
11004,
294,
428,
2690,
2245,
13,
50608,
50644,
467,
311,
445,
4930,
11,
309,
311,
1596,
4930,
13,
286,
2041,
8328,
952,
311,
935,
295,
577,
867,
295,
613,
50924,
50924,
2690,
712,
5245,
293,
577,
867,
295,
613,
8122,
366,
1944,
294,
4265,
1496,
5245,
11,
689,
411,
51224,
51224,
264,
2316,
2564,
307,
7633,
11,
291,
500,
380,
534,
458,
577,
411,
264,
1412,
307,
16146,
11,
498,
264,
1412,
307,
51444,
51444,
48704,
11,
721,
411,
300,
307,
4961,
11,
558,
30,
11164,
538,
22082,
3051,
13,
583,
412,
264,
917,
295,
264,
786,
11,
51772,
51812
]
},
{
"avg_logprob": -0.2660507675289184,
"compression_ratio": 1.7285714285714286,
"end": 1556.24,
"id": 238,
"no_speech_prob": 0.000051439827075228095,
"seek": 155368,
"start": 1553.68,
"temperature": 0,
"text": " basically in theory, it's very hard to defend.",
"tokens": [
50364,
1936,
294,
5261,
11,
309,
311,
588,
1152,
281,
8602,
13,
50492,
50648,
286,
478,
516,
281,
3012,
294,
257,
3456,
11,
457,
286,
445,
528,
281,
652,
472,
2871,
11,
597,
307,
300,
50780,
50780,
286,
519,
309,
311,
1021,
281,
519,
466,
1176,
42,
12683,
382,
445,
411,
472,
644,
295,
364,
4787,
17240,
1185,
13,
51020,
51020,
467,
311,
445,
411,
257,
2167,
2522,
295,
257,
17240,
1185,
11,
291,
458,
11,
498,
291,
500,
380,
5139,
428,
33149,
1562,
11,
51324,
51324,
558,
30,
1743,
291,
393,
652,
512,
6146,
294,
264,
1185,
300,
1177,
380,
3314,
2295,
561,
11,
51512,
51512,
457,
561,
360,
1333,
295,
6322,
8122,
13,
407,
286,
519,
2035,
321,
1715,
1228,
1176,
42,
12683,
11,
51704,
51704
]
},
{
"avg_logprob": -0.2660507675289184,
"compression_ratio": 1.7285714285714286,
"end": 1562,
"id": 239,
"no_speech_prob": 0.000051439827075228095,
"seek": 155368,
"start": 1559.3600000000001,
"temperature": 0,
"text": " I'm going to jump in a minute, but I just want to make one comment, which is that",
"tokens": [
50364,
1936,
294,
5261,
11,
309,
311,
588,
1152,
281,
8602,
13,
50492,
50648,
286,
478,
516,
281,
3012,
294,
257,
3456,
11,
457,
286,
445,
528,
281,
652,
472,
2871,
11,
597,
307,
300,
50780,
50780,
286,
519,
309,
311,
1021,
281,
519,
466,
1176,
42,
12683,
382,
445,
411,
472,
644,
295,
364,
4787,
17240,
1185,
13,
51020,
51020,
467,
311,
445,
411,
257,
2167,
2522,
295,
257,
17240,
1185,
11,
291,
458,
11,
498,
291,
500,
380,
5139,
428,
33149,
1562,
11,
51324,
51324,
558,
30,
1743,
291,
393,
652,
512,
6146,
294,
264,
1185,
300,
1177,
380,
3314,
2295,
561,
11,
51512,
51512,
457,
561,
360,
1333,
295,
6322,
8122,
13,
407,
286,
519,
2035,
321,
1715,
1228,
1176,
42,
12683,
11,
51704,
51704
]
},
{
"avg_logprob": -0.2660507675289184,
"compression_ratio": 1.7285714285714286,
"end": 1566.8,
"id": 240,
"no_speech_prob": 0.000051439827075228095,
"seek": 155368,
"start": 1562,
"temperature": 0,
"text": " I think it's important to think about ZKML as just like one part of an overall crypto system.",
"tokens": [
50364,
1936,
294,
5261,
11,
309,
311,
588,
1152,
281,
8602,
13,
50492,
50648,
286,
478,
516,
281,
3012,
294,
257,
3456,
11,
457,
286,
445,
528,
281,
652,
472,
2871,
11,
597,
307,
300,
50780,
50780,
286,
519,
309,
311,
1021,
281,
519,
466,
1176,
42,
12683,
382,
445,
411,
472,
644,
295,
364,
4787,
17240,
1185,
13,
51020,
51020,
467,
311,
445,
411,
257,
2167,
2522,
295,
257,
17240,
1185,
11,
291,
458,
11,
498,
291,
500,
380,
5139,
428,
33149,
1562,
11,
51324,
51324,
558,
30,
1743,
291,
393,
652,
512,
6146,
294,
264,
1185,
300,
1177,
380,
3314,
2295,
561,
11,
51512,
51512,
457,
561,
360,
1333,
295,
6322,
8122,
13,
407,
286,
519,
2035,
321,
1715,
1228,
1176,
42,
12683,
11,
51704,
51704
]
},
{
"avg_logprob": -0.2660507675289184,
"compression_ratio": 1.7285714285714286,
"end": 1572.88,
"id": 241,
"no_speech_prob": 0.000051439827075228095,
"seek": 155368,
"start": 1566.8,
"temperature": 0,
"text": " It's just like a single piece of a crypto system, you know, if you don't salt your passwords ever,",
"tokens": [
50364,
1936,
294,
5261,
11,
309,
311,
588,
1152,
281,
8602,
13,
50492,
50648,
286,
478,
516,
281,
3012,
294,
257,
3456,
11,
457,
286,
445,
528,
281,
652,
472,
2871,
11,
597,
307,
300,
50780,
50780,
286,
519,
309,
311,
1021,
281,
519,
466,
1176,
42,
12683,
382,
445,
411,
472,
644,
295,
364,
4787,
17240,
1185,
13,
51020,
51020,
467,
311,
445,
411,
257,
2167,
2522,
295,
257,
17240,
1185,
11,
291,
458,
11,
498,
291,
500,
380,
5139,
428,
33149,
1562,
11,
51324,
51324,
558,
30,
1743,
291,
393,
652,
512,
6146,
294,
264,
1185,
300,
1177,
380,
3314,
2295,
561,
11,
51512,
51512,
457,
561,
360,
1333,
295,
6322,
8122,
13,
407,
286,
519,
2035,
321,
1715,
1228,
1176,
42,
12683,
11,
51704,
51704
]
},
{
"avg_logprob": -0.2660507675289184,
"compression_ratio": 1.7285714285714286,
"end": 1576.64,
"id": 242,
"no_speech_prob": 0.000051439827075228095,
"seek": 155368,
"start": 1572.88,
"temperature": 0,
"text": " right? Like you can make some mistake in the system that doesn't rate low people,",
"tokens": [
50364,
1936,
294,
5261,
11,
309,
311,
588,
1152,
281,
8602,
13,
50492,
50648,
286,
478,
516,
281,
3012,
294,
257,
3456,
11,
457,
286,
445,
528,
281,
652,
472,
2871,
11,
597,
307,
300,
50780,
50780,
286,
519,
309,
311,
1021,
281,
519,
466,
1176,
42,
12683,
382,
445,
411,
472,
644,
295,
364,
4787,
17240,
1185,
13,
51020,
51020,
467,
311,
445,
411,
257,
2167,
2522,
295,
257,
17240,
1185,
11,
291,
458,
11,
498,
291,
500,
380,
5139,
428,
33149,
1562,
11,
51324,
51324,
558,
30,
1743,
291,
393,
652,
512,
6146,
294,
264,
1185,
300,
1177,
380,
3314,
2295,
561,
11,
51512,
51512,
457,
561,
360,
1333,
295,
6322,
8122,
13,
407,
286,
519,
2035,
321,
1715,
1228,
1176,
42,
12683,
11,
51704,
51704
]
},
{
"avg_logprob": -0.2660507675289184,
"compression_ratio": 1.7285714285714286,
"end": 1580.48,
"id": 243,
"no_speech_prob": 0.000051439827075228095,
"seek": 155368,
"start": 1576.64,
"temperature": 0,
"text": " but people do sort of obvious attacks. So I think whatever we design using ZKML,",
"tokens": [
50364,
1936,
294,
5261,
11,
309,
311,
588,
1152,
281,
8602,
13,
50492,
50648,
286,
478,
516,
281,
3012,
294,
257,
3456,
11,
457,
286,
445,
528,
281,
652,
472,
2871,
11,
597,
307,
300,
50780,
50780,
286,
519,
309,
311,
1021,
281,
519,
466,
1176,
42,
12683,
382,
445,
411,
472,
644,
295,
364,
4787,
17240,
1185,
13,
51020,
51020,
467,
311,
445,
411,
257,
2167,
2522,
295,
257,
17240,
1185,
11,
291,
458,
11,
498,
291,
500,
380,
5139,
428,
33149,
1562,
11,
51324,
51324,
558,
30,
1743,
291,
393,
652,
512,
6146,
294,
264,
1185,
300,
1177,
380,
3314,
2295,
561,
11,
51512,
51512,
457,
561,
360,
1333,
295,
6322,
8122,
13,
407,
286,
519,
2035,
321,
1715,
1228,
1176,
42,
12683,
11,
51704,
51704
]
},
{
"avg_logprob": -0.18637037896490716,
"compression_ratio": 1.4759615384615385,
"end": 1585.3600000000001,
"id": 244,
"no_speech_prob": 0.00014879336231388152,
"seek": 158048,
"start": 1580.48,
"temperature": 0,
"text": " especially for security, we have to think about the context of the overall crypto system and how",
"tokens": [
50364,
2318,
337,
3825,
11,
321,
362,
281,
519,
466,
264,
4319,
295,
264,
4787,
17240,
1185,
293,
577,
50608,
50608,
360,
321,
829,
264,
3755,
1214,
281,
652,
8122,
411,
341,
2252,
420,
6243,
13,
1044,
291,
337,
264,
50896,
50896,
5017,
13,
1033,
13,
51320,
51388,
5818,
3270,
13,
865,
11,
1338,
11,
321,
366,
484,
295,
264,
565,
11,
457,
611,
286,
478,
406,
988,
466,
577,
709,
286,
362,
337,
51776,
51776
]
},
{
"avg_logprob": -0.18637037896490716,
"compression_ratio": 1.4759615384615385,
"end": 1591.1200000000001,
"id": 245,
"no_speech_prob": 0.00014879336231388152,
"seek": 158048,
"start": 1585.3600000000001,
"temperature": 0,
"text": " do we put the pieces together to make attacks like this difficult or impossible. Thank you for the",
"tokens": [
50364,
2318,
337,
3825,
11,
321,
362,
281,
519,
466,
264,
4319,
295,
264,
4787,
17240,
1185,
293,
577,
50608,
50608,
360,
321,
829,
264,
3755,
1214,
281,
652,
8122,
411,
341,
2252,
420,
6243,
13,
1044,
291,
337,
264,
50896,
50896,
5017,
13,
1033,
13,
51320,
51388,
5818,
3270,
13,
865,
11,
1338,
11,
321,
366,
484,
295,
264,
565,
11,
457,
611,
286,
478,
406,
988,
466,
577,
709,
286,
362,
337,
51776,
51776
]
},
{
"avg_logprob": -0.18637037896490716,
"compression_ratio": 1.4759615384615385,
"end": 1599.6,
"id": 246,
"no_speech_prob": 0.00014879336231388152,
"seek": 158048,
"start": 1591.1200000000001,
"temperature": 0,
"text": " discussion. Okay.",
"tokens": [
50364,
2318,
337,
3825,
11,
321,
362,
281,
519,
466,
264,
4319,
295,
264,
4787,
17240,
1185,
293,
577,
50608,
50608,
360,
321,
829,
264,
3755,
1214,
281,
652,
8122,
411,
341,
2252,
420,
6243,
13,
1044,
291,
337,
264,
50896,
50896,
5017,
13,
1033,
13,
51320,
51388,
5818,
3270,
13,
865,
11,
1338,
11,
321,
366,
484,
295,
264,
565,
11,
457,
611,
286,
478,
406,
988,
466,
577,
709,
286,
362,
337,
51776,
51776
]
},
{
"avg_logprob": -0.18637037896490716,
"compression_ratio": 1.4759615384615385,
"end": 1608.72,
"id": 247,
"no_speech_prob": 0.00014879336231388152,
"seek": 158048,
"start": 1600.96,
"temperature": 0,
"text": " Mic drop. Yeah, yeah, we are out of the time, but also I'm not sure about how much I have for",
"tokens": [
50364,
2318,
337,
3825,
11,
321,
362,
281,
519,
466,
264,
4319,
295,
264,
4787,
17240,
1185,
293,
577,
50608,
50608,
360,
321,
829,
264,
3755,
1214,
281,
652,
8122,
411,
341,
2252,
420,
6243,
13,
1044,
291,
337,
264,
50896,
50896,
5017,
13,
1033,
13,
51320,
51388,
5818,
3270,
13,
865,
11,
1338,
11,
321,
366,
484,
295,
264,
565,
11,
457,
611,
286,
478,
406,
988,
466,
577,
709,
286,
362,
337,
51776,
51776
]
},
{
"avg_logprob": -0.14860777421431107,
"compression_ratio": 1.5363128491620113,
"end": 1616,
"id": 248,
"no_speech_prob": 0.0005418873042799532,
"seek": 160872,
"start": 1608.72,
"temperature": 0,
"text": " recording, not on my laptop, but via Jitsi. So I think that for this call, we should be very",
"tokens": [
50364,
6613,
11,
406,
322,
452,
10732,
11,
457,
5766,
508,
1208,
72,
13,
407,
286,
519,
300,
337,
341,
818,
11,
321,
820,
312,
588,
50728,
50728,
11239,
295,
264,
4787,
1874,
13,
286,
914,
11,
412,
1935,
286,
519,
300,
390,
370,
1577,
295,
2701,
13,
51100,
51160,
759,
291,
3986,
11,
457,
286,
519,
300,
291,
393,
1029,
611,
281,
264,
14889,
1342,
2269,
13,
286,
576,
959,
498,
321,
393,
51504,
51564
]
},
{
"avg_logprob": -0.14860777421431107,
"compression_ratio": 1.5363128491620113,
"end": 1623.44,
"id": 249,
"no_speech_prob": 0.0005418873042799532,
"seek": 160872,
"start": 1616,
"temperature": 0,
"text": " satisfied of the overall result. I mean, at least I think that was so full of content.",
"tokens": [
50364,
6613,
11,
406,
322,
452,
10732,
11,
457,
5766,
508,
1208,
72,
13,
407,
286,
519,
300,
337,
341,
818,
11,
321,
820,
312,
588,
50728,
50728,
11239,
295,
264,
4787,
1874,
13,
286,
914,
11,
412,
1935,
286,
519,
300,
390,
370,
1577,
295,
2701,
13,
51100,
51160,
759,
291,
3986,
11,
457,
286,
519,
300,
291,
393,
1029,
611,
281,
264,
14889,
1342,
2269,
13,
286,
576,
959,
498,
321,
393,
51504,
51564
]
},
{
"avg_logprob": -0.14860777421431107,
"compression_ratio": 1.5363128491620113,
"end": 1631.52,
"id": 250,
"no_speech_prob": 0.0005418873042799532,
"seek": 160872,
"start": 1624.64,
"temperature": 0,
"text": " If you agree, but I think that you can ask also to the Telegram channel. I would love if we can",
"tokens": [
50364,
6613,
11,
406,
322,
452,
10732,
11,
457,
5766,
508,
1208,
72,
13,
407,
286,
519,
300,
337,
341,
818,
11,
321,
820,
312,
588,
50728,
50728,
11239,
295,
264,
4787,
1874,
13,
286,
914,
11,
412,
1935,
286,
519,
300,
390,
370,
1577,
295,
2701,
13,
51100,
51160,
759,
291,
3986,
11,
457,
286,
519,
300,
291,
393,
1029,
611,
281,
264,
14889,
1342,
2269,
13,
286,
576,
959,
498,
321,
393,
51504,
51564
]
},
{
"avg_logprob": -0.1811329783225546,
"compression_ratio": 1.6343612334801763,
"end": 1638.24,
"id": 251,
"no_speech_prob": 0.0008157803677022457,
"seek": 163152,
"start": 1631.52,
"temperature": 0,
"text": " do this frequently, maybe not every week, but maybe bi-weekly. So every two weeks.",
"tokens": [
50364,
360,
341,
10374,
11,
1310,
406,
633,
1243,
11,
457,
1310,
3228,
12,
23188,
356,
13,
407,
633,
732,
3259,
13,
50700,
50752,
759,
341,
3263,
665,
11,
321,
393,
1393,
337,
264,
11,
291,
458,
11,
337,
264,
2808,
295,
7674,
420,
264,
1150,
1243,
51024,
51024,
295,
7674,
13,
1144,
291,
411,
309,
30,
14576,
665,
281,
385,
13,
865,
11,
286,
576,
3402,
611,
2891,
365,
257,
3098,
51412,
51412,
295,
560,
2635,
370,
321,
458,
437,
1518,
311,
884,
293,
1364,
322,
13,
7021,
13,
7021,
13,
286,
632,
512,
51740,
51740
]
},
{
"avg_logprob": -0.1811329783225546,
"compression_ratio": 1.6343612334801763,
"end": 1644.72,
"id": 252,
"no_speech_prob": 0.0008157803677022457,
"seek": 163152,
"start": 1639.28,
"temperature": 0,
"text": " If this sounds good, we can plan for the, you know, for the middle of November or the second week",
"tokens": [
50364,
360,
341,
10374,
11,
1310,
406,
633,
1243,
11,
457,
1310,
3228,
12,
23188,
356,
13,
407,
633,
732,
3259,
13,
50700,
50752,
759,
341,
3263,
665,
11,
321,
393,
1393,
337,
264,
11,
291,
458,
11,
337,
264,
2808,
295,
7674,
420,
264,
1150,
1243,
51024,
51024,
295,
7674,
13,
1144,
291,
411,
309,
30,
14576,
665,
281,
385,
13,
865,
11,
286,
576,
3402,
611,
2891,
365,
257,
3098,
51412,
51412,
295,
560,
2635,
370,
321,
458,
437,
1518,
311,
884,
293,
1364,
322,
13,
7021,
13,
7021,
13,
286,
632,
512,
51740,
51740
]
},
{
"avg_logprob": -0.1811329783225546,
"compression_ratio": 1.6343612334801763,
"end": 1652.48,
"id": 253,
"no_speech_prob": 0.0008157803677022457,
"seek": 163152,
"start": 1644.72,
"temperature": 0,
"text": " of November. Do you like it? Sounds good to me. Yeah, I would suggest also starting with a round",
"tokens": [
50364,
360,
341,
10374,
11,
1310,
406,
633,
1243,
11,
457,
1310,
3228,
12,
23188,
356,
13,
407,
633,
732,
3259,
13,
50700,
50752,
759,
341,
3263,
665,
11,
321,
393,
1393,
337,
264,
11,
291,
458,
11,
337,
264,
2808,
295,
7674,
420,
264,
1150,
1243,
51024,
51024,
295,
7674,
13,
1144,
291,
411,
309,
30,
14576,
665,
281,
385,
13,
865,
11,
286,
576,
3402,
611,
2891,
365,
257,
3098,
51412,
51412,
295,
560,
2635,
370,
321,
458,
437,
1518,
311,
884,
293,
1364,
322,
13,
7021,
13,
7021,
13,
286,
632,
512,
51740,
51740
]
},
{
"avg_logprob": -0.1811329783225546,
"compression_ratio": 1.6343612334801763,
"end": 1659.04,
"id": 254,
"no_speech_prob": 0.0008157803677022457,
"seek": 163152,
"start": 1652.48,
"temperature": 0,
"text": " of intros so we know what everyone's doing and working on. Absolutely. Absolutely. I had some",
"tokens": [
50364,
360,
341,
10374,
11,
1310,
406,
633,
1243,
11,
457,
1310,
3228,
12,
23188,
356,
13,
407,
633,
732,
3259,
13,
50700,
50752,
759,
341,
3263,
665,
11,
321,
393,
1393,
337,
264,
11,
291,
458,
11,
337,
264,
2808,
295,
7674,
420,
264,
1150,
1243,
51024,
51024,
295,
7674,
13,
1144,
291,
411,
309,
30,
14576,
665,
281,
385,
13,
865,
11,
286,
576,
3402,
611,
2891,
365,
257,
3098,
51412,
51412,
295,
560,
2635,
370,
321,
458,
437,
1518,
311,
884,
293,
1364,
322,
13,
7021,
13,
7021,
13,
286,
632,
512,
51740,
51740
]
},
{
"avg_logprob": -0.25052209096412137,
"compression_ratio": 1.53125,
"end": 1665.92,
"id": 255,
"no_speech_prob": 0.0018566870130598545,
"seek": 165904,
"start": 1659.04,
"temperature": 0,
"text": " problem with the webcam, so I have to get back and forth, but yeah, next time we can dedicate half of",
"tokens": [
50364,
1154,
365,
264,
39490,
11,
370,
286,
362,
281,
483,
646,
293,
5220,
11,
457,
1338,
11,
958,
565,
321,
393,
30718,
1922,
295,
50708,
50708,
264,
818,
11,
291,
458,
11,
281,
458,
1101,
1184,
661,
13,
865,
11,
2138,
13,
2555,
13365,
264,
5570,
300,
51132,
51132,
9114,
11875,
433,
2073,
294,
264,
5081,
365,
6342,
370,
321,
393,
11374,
294,
264,
3855,
295,
264,
960,
300,
51632,
51632
]
},
{
"avg_logprob": -0.25052209096412137,
"compression_ratio": 1.53125,
"end": 1674.3999999999999,
"id": 256,
"no_speech_prob": 0.0018566870130598545,
"seek": 165904,
"start": 1665.92,
"temperature": 0,
"text": " the call, you know, to know better each other. Yeah, definitely. Please integrate the notes that",
"tokens": [
50364,
1154,
365,
264,
39490,
11,
370,
286,
362,
281,
483,
646,
293,
5220,
11,
457,
1338,
11,
958,
565,
321,
393,
30718,
1922,
295,
50708,
50708,
264,
818,
11,
291,
458,
11,
281,
458,
1101,
1184,
661,
13,
865,
11,
2138,
13,
2555,
13365,
264,
5570,
300,
51132,
51132,
9114,
11875,
433,
2073,
294,
264,
5081,
365,
6342,
370,
321,
393,
11374,
294,
264,
3855,
295,
264,
960,
300,
51632,
51632
]
},
{
"avg_logprob": -0.25052209096412137,
"compression_ratio": 1.53125,
"end": 1684.3999999999999,
"id": 257,
"no_speech_prob": 0.0018566870130598545,
"seek": 165904,
"start": 1674.3999999999999,
"temperature": 0,
"text": " DC Builders share in the chat with yours so we can publish in the description of the video that",
"tokens": [
50364,
1154,
365,
264,
39490,
11,
370,
286,
362,
281,
483,
646,
293,
5220,
11,
457,
1338,
11,
958,
565,
321,
393,
30718,
1922,
295,
50708,
50708,
264,
818,
11,
291,
458,
11,
281,
458,
1101,
1184,
661,
13,
865,
11,
2138,
13,
2555,
13365,
264,
5570,
300,
51132,
51132,
9114,
11875,
433,
2073,
294,
264,
5081,
365,
6342,
370,
321,
393,
11374,
294,
264,
3855,
295,
264,
960,
300,
51632,
51632
]
},
{
"avg_logprob": -0.3189210891723633,
"compression_ratio": 1.2,
"end": 1690.4,
"id": 258,
"no_speech_prob": 0.0020998362451791763,
"seek": 168440,
"start": 1684.4,
"temperature": 0,
"text": " we are going to publish on YouTube and also in some repository for, you know, others to see. Okay.",
"tokens": [
50364,
321,
366,
516,
281,
11374,
322,
3088,
293,
611,
294,
512,
25841,
337,
11,
291,
458,
11,
2357,
281,
536,
13,
1033,
13,
50664,
50864,
8561,
13,
407,
718,
385,
584,
1590,
264,
6613,
13,
51056
]
},
{
"avg_logprob": -0.42064949444362093,
"compression_ratio": 0.84,
"end": 1712.4,
"id": 259,
"no_speech_prob": 0.0018217681208625436,
"seek": 169040,
"start": 1690.4,
"temperature": 0,
"text": " Cool. So let me say stop in the recording.",
"tokens": [
50364,
8561,
13,
407,
718,
385,
584,
1590,
294,
264,
6613,
13,
51464
]
}
]1 00:00:02,640 --> 00:00:03,140 Okay. 2 00:00:05,520 --> 00:00:10,960 Doxxed. Everyone. Cool. Please, Tyler, go ahead. Thank you, guys. 3 00:00:12,080 --> 00:00:18,160 Yeah, thanks. I'm probably the least intelligent person on this call, 4 00:00:19,840 --> 00:00:25,440 but I know the Modulus Lab guys. I think we're like 40 feet away from each other right now. 5 00:00:25,440 --> 00:00:33,840 So I built and run the program that they went through. They landed on Modulus Labs. But I guess 6 00:00:33,840 --> 00:00:44,000 one thing I'm trying to wrap my head around in this space is, like, are there certain things that 7 00:00:44,000 --> 00:00:52,160 ZKML can enable that don't necessarily provide the better back-end experience or where, like, 8 00:00:52,160 --> 00:00:59,440 oh, the tech makes this thing better, but where it provides, like, 10x better, like, 9 00:00:59,440 --> 00:01:07,280 end-user or, like, consumer experience? I guess this is just, like, a problem I see across crypto, 10 00:01:07,280 --> 00:01:12,960 where, like, I think all of us see the promise of crypto, but we don't really see it being readily 11 00:01:12,960 --> 00:01:22,320 adopted. And so, yeah, one of the things I'm just trying to spend time on is where are the places 12 00:01:22,320 --> 00:01:32,480 where this can be applied that, you know, like, fundamentally change the way that someone would 13 00:01:32,480 --> 00:01:39,520 be able to interact with this ecosystem? I don't know if there's a good answer for that. 14 00:01:39,520 --> 00:01:45,520 I have to jump in because this is actually why I started working on this was because I think it is 15 00:01:45,520 --> 00:01:50,240 the kind of the right solution. We're looking for a way to own digital assets and to control 16 00:01:50,240 --> 00:01:54,640 smart contracts or control smart wallets, basically, control our whatever we have activity 17 00:01:54,640 --> 00:01:59,680 on chain using kind of more traditional notions of ownership that anyone can understand, right? 18 00:01:59,680 --> 00:02:03,920 That are more like about how we're embedded in society, who we are, where we are, all the things 19 00:02:03,920 --> 00:02:11,120 that we could do, right? So after we've met, I often give the example after we've met and talked 20 00:02:11,120 --> 00:02:18,400 a bit, I could tell you, I could say, please send 10 ETH to Emiliano, right? And you'd be able to do 21 00:02:18,400 --> 00:02:23,760 that fairly well, you know, maybe not on a video call with deepfakes and stuff, but in person, you 22 00:02:23,760 --> 00:02:28,320 could do it, right? And so why can't we do that with, you know, why do we have to use wallets and 23 00:02:28,320 --> 00:02:33,440 secrets and everything? And that's just a purely technical problem. And if you think about this 24 00:02:33,440 --> 00:02:38,400 model, like Celeste was talking about earlier, an ensemble of models, ZKML models that are measuring 25 00:02:38,400 --> 00:02:43,760 everything about you, sort of where you are, your devices, your, you know, your image and voice and 26 00:02:43,760 --> 00:02:48,880 your location, history and all this kind of stuff, you could eventually imagine having kind of 10,000 27 00:02:48,880 --> 00:02:53,520 factor authentication, where you really have a very high degree of confidence that this is this 28 00:02:53,520 --> 00:02:58,400 person, and this is the intent they've expressed. And then you execute that intent. And that could 29 00:02:58,400 --> 00:03:06,320 all happen in sort of an automated way. So I think that that would be the huge win for the, for the 30 00:03:06,320 --> 00:03:11,520 consumer and for ease of use. But it's probably very far off in terms of the technology getting 31 00:03:11,520 --> 00:03:16,480 there. And so I think about the problem we have to solve to get there is what are all of the smaller 32 00:03:16,480 --> 00:03:20,880 problems we can solve along the way? How do we get that flywheel spinning, so that we build the 33 00:03:20,880 --> 00:03:24,960 infrastructure and the models and everything else to be able to deliver that kind of experience 34 00:03:24,960 --> 00:03:33,680 eventually? Yeah, I mean, one of the companies echoing, I think Jason's point, like one of the 35 00:03:33,680 --> 00:03:37,680 companies that I was personally very excited about, like five years ago, or something, 36 00:03:37,680 --> 00:03:43,760 was this very cool tech company called Unifi ID. And their whole point was like, on your smartphone, 37 00:03:43,760 --> 00:03:48,960 they would track all of these kind of like personal data points that Jason was talking about. 38 00:03:48,960 --> 00:03:54,320 And in their TechCrunch demo, I think on stage, they had one co founder hand the phone to his 39 00:03:54,320 --> 00:03:58,880 other co founder, the other co founder walked around with the phone for like a couple seconds, 40 00:03:58,880 --> 00:04:04,240 tried to log in, and it said blocked by Unifi ID, like clearly your gate is not the same as 41 00:04:04,240 --> 00:04:11,360 that of the original owner, which is super, super neat. Probably the challenge there is then 42 00:04:11,360 --> 00:04:17,520 getting the models to run vocally, because the alternative then is you ping a server, 43 00:04:17,520 --> 00:04:22,480 which has all these like biometric models stored on it. And the server has to talk to the chain, 44 00:04:22,480 --> 00:04:27,040 and you have to talk to the server. And there's all this indirection. So the ideal here would 45 00:04:27,040 --> 00:04:31,680 just be you have all the models on phone, and you can run the models, you can run proofs for 46 00:04:31,680 --> 00:04:37,520 the models. And accompanied with every transaction is basically a proof of saying like, yes, this was 47 00:04:37,520 --> 00:04:42,400 actually sent by me. And by the way, here's what I want to do, like no more need for a secret key, 48 00:04:42,400 --> 00:04:58,000 no more need for any of that. Yeah. Corin, folks, you want to add something to that? 49 00:05:03,520 --> 00:05:10,240 You know what, following up the Jonathan, latest statement in the chat, 50 00:05:10,240 --> 00:05:16,960 we'd love to know if there is some, but this is more for machine learning than ZK, a way to 51 00:05:18,400 --> 00:05:27,600 understand if two or different Twitter profiles are related to the same person, you know, and 52 00:05:28,240 --> 00:05:37,520 using ZKML or this kind of mix to provide some sort of CBL resistance for, especially recently, 53 00:05:37,520 --> 00:05:45,120 there was a discussion of, I mean, we already knew that someone is doing that for, you know, 54 00:05:47,680 --> 00:05:54,320 doing similar attacks for airdrops. You discover probably with Uniswap when you use burner account 55 00:05:54,320 --> 00:06:00,320 that basically you receive many drops, you know, and so someone is trying to industrialize it. 56 00:06:00,320 --> 00:06:05,760 So I think that for, you know, there are many attempts like proof of humanity that do a video 57 00:06:05,760 --> 00:06:11,440 recording with your wallet address, you've seen a transaction chain, you know, there are many 58 00:06:11,440 --> 00:06:18,560 approaches that we can debate about the security assumptions and many, many others, okay, 59 00:06:19,760 --> 00:06:25,760 that someone is showing off. But yeah, the question for me is in general, let's say, despite 60 00:06:25,760 --> 00:06:35,680 the specific protocol, whatever, taking an approach of working backwards to see which could be the 61 00:06:36,160 --> 00:06:40,960 use case, how do you think that we can apply this technology to those things? 62 00:06:43,440 --> 00:06:50,560 I mean, this is basically what we're doing. One of the hard problems here is linking it back to 63 00:06:50,560 --> 00:06:57,280 reality. So getting some sort of trusted, you need some sort of trusted Oracle until a physicist 64 00:06:57,280 --> 00:07:04,880 figured out how to implement Diffie-Hellman in the laws of nature. And the way we do that is by the 65 00:07:04,880 --> 00:07:10,400 shiny little orb that the DZ builder just showed on camera, you can still see it behind him on a 66 00:07:10,400 --> 00:07:16,640 stand. So once you accept that, and once you get over that hurdle of, okay, we need some form of 67 00:07:16,640 --> 00:07:22,160 trusted hardware to hook into reality, then you move the problem to okay, now we have authentic 68 00:07:22,160 --> 00:07:28,480 biometric data. And then you can do all sorts of interesting things using CKML to process that data 69 00:07:29,600 --> 00:07:35,040 in a way that gives you self custody and is decentralized and trustless and all those nice 70 00:07:35,040 --> 00:07:42,320 properties that we want. And this is actually how I got to the field of like, ZKML in the first 71 00:07:42,320 --> 00:07:47,760 place, it was basically right now the orb does the biometric processing and trusted hardware. And my 72 00:07:47,760 --> 00:07:52,160 question was, okay, what if we want to update the model? Does now everyone need to go back to this 73 00:07:52,160 --> 00:08:00,880 orb? And we can solve that with CKML by basically just signing the raw data and doing all the 74 00:08:00,880 --> 00:08:10,720 processing in the user's wallet itself, which solves a huge UX issue for us. And then it's 75 00:08:10,720 --> 00:08:15,520 turns out that this actually opens a number of other interesting doors. If there is a way for 76 00:08:15,520 --> 00:08:24,480 example, to get reasonably secure authenticated data out of a smartphone camera, then we can do 77 00:08:24,480 --> 00:08:29,120 things like, oh, make sure that the user who is about to make a big transaction using this wallet 78 00:08:29,120 --> 00:08:38,800 is that is like, facially matches the user who originally created this wallet. So it has a number 79 00:08:38,800 --> 00:08:46,480 of like, security advantages and UX advantages we can achieve here. I think that's, and then there's 80 00:08:46,480 --> 00:08:52,000 like a bunch of other things that we're exploring where you can match biometrics across different 81 00:08:52,000 --> 00:08:59,120 things like, you could match facial biometrics from the orb with facial biometrics that are 82 00:08:59,120 --> 00:09:05,120 stored on government digital documents, for example. And it opens up a number of interesting 83 00:09:05,120 --> 00:09:11,120 use cases. Once you have access to authenticated sensor data, basically. 84 00:09:13,520 --> 00:09:20,560 Do you think that, I mean, trusting hardware, trusting that devices, you know, 85 00:09:22,240 --> 00:09:26,240 could be pretty hard, you know, could be a barrier. Do you think that initiatives that 86 00:09:27,040 --> 00:09:35,040 try to create some sort of standard or maybe some open hardware design could be a barrier? 87 00:09:35,040 --> 00:09:42,720 Could have to lower. So basically having people that, you know, build, you know, different 88 00:09:42,720 --> 00:09:48,560 version of the same trusted device, and you know, effectively what it does, or what could be the 89 00:09:48,560 --> 00:09:56,320 architecture could solve or mitigate some of the, you know, trust assumptions that a user, I mean, 90 00:09:57,360 --> 00:10:04,560 may face on, you know what, I have to trust. I mean, we already trust the hyphen when we unlock 91 00:10:04,560 --> 00:10:11,440 it. So it's just, you know, a matter of, you know, explaining that, but obviously, 92 00:10:12,000 --> 00:10:18,080 having some open standard would, you know, help in this sense, at least this is my idea. 93 00:10:18,080 --> 00:10:26,400 I mean, we are opening the standards. The problem is not so much that it's hard for people to 94 00:10:26,400 --> 00:10:31,840 implement it. The problem is that you don't know that they actually implemented it correctly. Like, 95 00:10:31,840 --> 00:10:36,880 let's say you have an authenticating photo camera, and you create an open standard for that. 96 00:10:38,080 --> 00:10:45,440 Let's say a dumb thing where you just like, let's say, sign a jpeg using a ECDSA signature. 97 00:10:45,440 --> 00:10:51,440 Now you want to create some sort of public repository of all the like known correct cameras. 98 00:10:51,440 --> 00:10:56,400 But how do you gatekeep this repository? Like, how do you maintain a list of known correct public 99 00:10:56,400 --> 00:11:03,760 keys? Yeah, I'll defend someone from... I mean, I understood. Yeah, you cannot attest that I did 100 00:11:03,760 --> 00:11:09,600 the physical devices, you know, I mean, yeah, we are, you know, like the Eisenberg, you know, 101 00:11:09,600 --> 00:11:16,160 microscope, basically, where the wave is collapsing in my eye, in my mind, whatever. So that's the 102 00:11:16,160 --> 00:11:25,200 same problem that you have in other. But, you know, we have some sort of, you know, already 103 00:11:25,200 --> 00:11:32,080 some scheme about, you know, game theory for all of course, where we have a decentralized network of 104 00:11:33,040 --> 00:11:41,520 people, entities, agents, that maybe have some skin in the game to, you know, to do a test 105 00:11:41,520 --> 00:11:48,480 that truthfully, you know, something. It isn't something that could solve. Or just traditional 106 00:11:48,480 --> 00:11:53,120 traditional trust models, like he just said, we trust Apple to do this stuff correctly when 107 00:11:53,120 --> 00:12:00,640 they do it. Like we trust, like if the iPhone through some hoops can produce authenticated 108 00:12:00,640 --> 00:12:06,800 images, then we would trust this to be correct. Because Apple is too big to like mess up a feature 109 00:12:06,800 --> 00:12:13,280 like that. Similarly, I think right now, either Nikon or Sony, I forget whom, but they also dabble 110 00:12:13,280 --> 00:12:18,640 in authenticating cameras. And for the web3 use case, we trust them to do that correctly, 111 00:12:18,640 --> 00:12:24,560 because we basically trust people at Sony and Nikon to not even know what web3 is, let alone 112 00:12:24,560 --> 00:12:30,160 build hardware exploits for their benefit. Because right now, the opportunity is not there. 113 00:12:30,160 --> 00:12:36,800 Problem is that as soon as the opportunities become there, and things do get big enough, then 114 00:12:37,520 --> 00:12:43,760 people will actually infiltrate supply chains to mess around with this. This is, for example, 115 00:12:43,760 --> 00:12:49,440 what is happening with decentralized exchanges, people literally apply for jobs at Verizon and 116 00:12:49,440 --> 00:12:55,440 T-Mobile, just so they get the authorization to do SIM swaps just so they can take over accounts 117 00:12:55,440 --> 00:13:02,400 at large centralized exchanges using their two factor authentication. Similar problems. Well, 118 00:13:03,200 --> 00:13:08,480 the same thing I would imagine happens once we go to trusted sensors, like initially, 119 00:13:08,480 --> 00:13:13,280 it will be all happy because the use case is so small that we don't have to worry about this. And 120 00:13:13,280 --> 00:13:20,880 then the value, like assets on the management, so to say, will become large enough that it becomes 121 00:13:20,880 --> 00:13:30,160 a viable supply chain attack. Yeah, but we can spread the risk across. If we are able to 122 00:13:31,280 --> 00:13:40,080 test assessing in different ways without concentrating on a single point, basically, 123 00:13:40,080 --> 00:13:49,600 the problem for SIM swap is that single accounts owes too much money that becomes very feasible to 124 00:13:49,600 --> 00:13:54,160 do that kind of attack. But if we are able to have, I don't remember who was saying, but 125 00:13:55,120 --> 00:14:04,640 multiple ways, some sort of multiple factor for authentication that could be just in the individual, 126 00:14:04,640 --> 00:14:12,240 you know, web, an SMS or whatever, but also via schemes like social recovery or social 127 00:14:12,240 --> 00:14:18,240 attestation. I mean, I know that Emiliano has that face, not because Emiliano has just a camera, 128 00:14:18,240 --> 00:14:25,600 but because Emiliano was in a picture with another guy that I trust or in a public event or whatever. 129 00:14:26,480 --> 00:14:34,480 So I think that probably if we have a framework where we can have all those multiple sources of 130 00:14:34,480 --> 00:14:44,480 data with the respective way for saying that something is true, it could work somehow, 131 00:14:45,280 --> 00:14:53,920 rather than specializing in specific technology. This is when you get to the end game of basically 132 00:14:53,920 --> 00:14:58,880 just doing full on by easy and risk analysis of like, hey, this person is claiming this 133 00:14:58,880 --> 00:15:06,480 authorization with this evidence, like, what is the probability that this is incorrect? And then 134 00:15:06,480 --> 00:15:13,040 you say, well, for small transactions, I'm willing to tolerate like a 1% false positive rate. And 135 00:15:13,840 --> 00:15:18,320 I'm even willing to like put my money on the line and having to reimburse this person. 136 00:15:18,320 --> 00:15:22,880 Whereas for very large transactions, you want maybe higher assurances, and you can optimize 137 00:15:22,880 --> 00:15:28,560 expected value and blah, blah. This is basically what big banks are currently doing, which is why 138 00:15:28,560 --> 00:15:33,360 you can get away with so many crappy things that the banks are doing, because they have these risk 139 00:15:33,360 --> 00:15:40,240 models and they just know what the trade offs are. Same thing with Google. And it's like CAPTCHA, 140 00:15:40,240 --> 00:15:44,800 where like sometimes it's a simple checkbox, sometimes you need to go through more hoops. 141 00:15:44,800 --> 00:15:48,880 It's never really clear why and how, but it's just based on their back end risk models. 142 00:15:49,680 --> 00:15:54,640 The same thing is true for insurance as well, right? Like people have a higher burden of proof. 143 00:15:54,640 --> 00:16:00,240 It's a more substantial claim, which is also like, we talk about end game, like some of this, 144 00:16:01,440 --> 00:16:08,960 wallet recovery stuff, as well as verifying insurance claims is somewhat like the end game 145 00:16:08,960 --> 00:16:14,240 for some of these models as well. I mean, just massive, massive business that can't exist on 146 00:16:14,240 --> 00:16:25,760 chain because there's no good verification of claims, but ML could be one tool. Cool. 147 00:16:27,840 --> 00:16:34,480 We just have just for the sake of time, five minutes. I think that we can stick over 148 00:16:34,480 --> 00:16:43,520 one topic basically or some follow up question to this. 149 00:16:45,520 --> 00:16:52,880 Yeah, maybe a question to the group here. Does anyone have a good overview article or just a 150 00:16:52,880 --> 00:17:01,760 graphic even of all the different types of ways we can do encrypted machine learning? So thinking 151 00:17:01,760 --> 00:17:10,000 about a mapping of FHE, of federated learning, differential privacy, and including ZKML and all 152 00:17:10,000 --> 00:17:16,560 its ways and forms. Is there a good all encompassing guide for this anywhere? 153 00:17:19,360 --> 00:17:24,000 Or should we put it together? I don't think there's an indicator of all those fields. I know that there's 154 00:17:24,000 --> 00:17:30,560 some FHE stuff. I've put together a guide called also ZKML, which is like, I could create ZKML stuff. 155 00:17:30,560 --> 00:17:38,080 I haven't seen anything about differential privacy. I haven't looked into that topic, 156 00:17:38,080 --> 00:17:44,080 but I'm pretty sure that you can just create a doc that just encompasses all of these resources together. 157 00:17:45,840 --> 00:17:49,600 Yeah, okay. I just created a lot of interesting combinations that you can where you mix 158 00:17:49,600 --> 00:17:54,080 two or three of these next generation technologies and you get generally some new 159 00:17:54,640 --> 00:17:59,280 new capability that you can't accomplish with either of them. 160 00:17:59,280 --> 00:18:03,040 So I think there'll probably be a lot of interesting things that come out that combine 161 00:18:03,040 --> 00:18:09,440 differential privacy and ZK or FHE or whatever. Pick any two and there's some new 162 00:18:09,440 --> 00:18:15,520 new function you can compute. It's really cool, but the computational overhead you create from this 163 00:18:16,240 --> 00:18:22,800 when you combine technologies like this is pretty big. Like 10,000 fold computational overhead 164 00:18:22,800 --> 00:18:28,960 over the computation itself, which is fun. But computers are so fast and we can't spend it only on 165 00:18:28,960 --> 00:18:35,520 your JavaScript animation. Have you heard that the actual JavaScript ecosystem is actually 166 00:18:35,520 --> 00:18:41,280 getting better now? You're using like Rust to go pack. It's just like new stuff. It's really fun. 167 00:18:41,280 --> 00:18:46,400 It's getting better. It used to be worse. The disgusted look on Ryan's face. 168 00:18:49,760 --> 00:18:54,800 By the way, DC builder, did you get to talk to the guy from FedML.ai? 169 00:18:54,800 --> 00:19:00,720 FedML, I don't think I've heard about that. 170 00:19:00,720 --> 00:19:05,920 I thought I connected you guys. Yeah, I posted it on the chat. 171 00:19:05,920 --> 00:19:07,120 Yeah, thanks. 172 00:19:07,120 --> 00:19:12,800 So it's a platform for doing federated machine learning. Basically, they created a bunch of 173 00:19:12,800 --> 00:19:28,880 primitives. Yeah. And they're also racing a pretty big seed or series A. 174 00:19:28,880 --> 00:19:35,200 We can talk about privacy maybe. I learned some terrifying facts about large language models in 175 00:19:35,200 --> 00:19:46,000 this ZKL knowledge podcast, with Florian Tramier, about how you get basically like private data 176 00:19:46,000 --> 00:19:53,120 randomly leaked through GPT-3 or GPT-2 that I had no idea about. Just like someone accidentally 177 00:19:53,120 --> 00:20:02,160 posted there or got, I guess, doxxed online and then the training data just picked it up and the 178 00:20:02,160 --> 00:20:09,920 model didn't transform it. So some people just included in the output of GPT-2 with their full 179 00:20:09,920 --> 00:20:21,920 data. I guess I can speak. The way these, at least the early kind of like RNN LSTM-based models work 180 00:20:21,920 --> 00:20:28,400 is that for each input token, you are obviously keeping track of the state of the sentence or 181 00:20:28,400 --> 00:20:34,880 your context. But basically, you are trying to output the greatest probability next token based 182 00:20:34,880 --> 00:20:39,920 on your context so far. And you can put attention over this if you have only attention that you have 183 00:20:39,920 --> 00:20:46,480 a transformer. But the idea here is that if there are very, very unique sequences in the data set, 184 00:20:46,480 --> 00:20:52,400 and if your like model dimensionality, so like your hidden state dimensionality is really high, 185 00:20:52,400 --> 00:20:56,560 then what ends up happening is the model quite literally just memorizes because it says, 186 00:20:56,560 --> 00:21:01,360 yeah, after this particular sequence of tokens, the only possible next token that I've seen in 187 00:21:01,360 --> 00:21:06,480 the whole database, which is not going to give me infinite loss basically, is this next token. 188 00:21:06,480 --> 00:21:13,360 And so it just spits out that memorized sequence. This kind of problem seems to be possibly 189 00:21:13,360 --> 00:21:19,200 inherent to just high dimensional model representations in general. I think there's 190 00:21:19,200 --> 00:21:25,200 been some work done on like trying to do robustness and like lower bounds on, I guess, 191 00:21:25,200 --> 00:21:30,080 like error or something like that. I'm not too familiar with the literature, but it does seem 192 00:21:30,080 --> 00:21:36,160 like for these like kind of deterministic complex representation models, memorization might be like 193 00:21:36,160 --> 00:21:41,440 an inherent problem. And also there's the whole issue with like querying, right, that Florian was 194 00:21:41,440 --> 00:21:46,000 talking about, in which you can almost like reconstruct model weights for smaller models 195 00:21:46,000 --> 00:21:49,280 by just querying the model enough times if you know the structure of the model. 196 00:21:49,280 --> 00:21:57,520 All right, thanks. That was very helpful. I had no idea how it works in particular, 197 00:21:57,520 --> 00:21:59,520 but that was explanatory. 198 00:22:07,520 --> 00:22:09,520 Wish I could make better otter sounds. 199 00:22:12,480 --> 00:22:16,800 Yeah, and then I guess even on the topic of like security with respect to weights, 200 00:22:16,800 --> 00:22:21,680 I mean, as Jason has been alluding to, I think multiple times now, like adversarial attacks are 201 00:22:21,680 --> 00:22:27,200 really, really bad in the sense that especially for image adversarial attacks or audio adversarial 202 00:22:27,200 --> 00:22:33,120 attacks, you can perturb the input signal by an imperceptible amount to a human and have the 203 00:22:33,120 --> 00:22:39,280 model output with like 99.8% confidence, just a totally different classification class, or have 204 00:22:39,280 --> 00:22:45,760 it regress basically whatever number you want it to output. And this is basically a consequence of 205 00:22:45,760 --> 00:22:50,560 the fact that models are differentiable and you can differentiate with respect to the outputs in 206 00:22:50,560 --> 00:22:55,520 a certain kind of, we call it like projected gradient descent way, to kind of minimize the 207 00:22:55,520 --> 00:23:00,880 impact, like the human perceptible impact on the input while still changing the output class. 208 00:23:01,680 --> 00:23:07,040 And so, yeah, I guess the point is in this case, if you know the model weights or if you have a 209 00:23:07,040 --> 00:23:11,840 good approximation of the model weights, you can conduct a white box attack, which means locally 210 00:23:11,840 --> 00:23:17,360 you just run the attack and then you are guaranteed basically that the model will accept it. 211 00:23:17,360 --> 00:23:23,920 And it's one of the things where it's kind of a double edged sword because ZKML can fix this 212 00:23:23,920 --> 00:23:28,160 problem by letting you hide the weights. It's not like a perfect bullet against adversarial attacks, 213 00:23:28,160 --> 00:23:33,840 but it helps a lot. But then you lose out on kind of the web free ethos of anyone being able to run 214 00:23:33,840 --> 00:23:38,400 the model. So it's, you can add the weights, but then you need to centralize the operator, 215 00:23:38,400 --> 00:23:42,000 or you cannot hide the weights, but then you have to deal with adversarial effects. 216 00:23:42,000 --> 00:23:49,600 So like one thought in this direction might even to, I guess, have like non-deterministic models or 217 00:23:49,600 --> 00:23:55,600 maybe models that aren't so easily back propped through, right? So like GANs, for example, have 218 00:23:55,600 --> 00:24:00,400 a sampling step, which you can, I guess, approximate in some sense with like some smooth function. 219 00:24:01,120 --> 00:24:06,240 But in general, if we like put enough discontinuities or train models where 220 00:24:06,240 --> 00:24:11,920 it's not so straightforward, I guess, like to go from output back to, or to go from changes in input 221 00:24:11,920 --> 00:24:16,800 or changes in intermediate to output, then showing the weights might actually be okay. 222 00:24:16,800 --> 00:24:25,920 Are there techniques to make the white bucket tax more costly to where like the profitability is 223 00:24:25,920 --> 00:24:32,160 reduced or where it's sort of a moving target where it would, you would have to continually 224 00:24:32,160 --> 00:24:35,200 recompute a new attack vector or a new attack strategy? 225 00:24:36,960 --> 00:24:44,000 Yeah, so in general, computing an attacked input with some of those state-of-the-art attacks, so 226 00:24:44,000 --> 00:24:48,080 based on what I remember, at least there's this really, really strong attack called Carlini-Wagner 227 00:24:48,080 --> 00:24:52,800 takes like a couple thousand basically like gradient descent iterations. So computing the 228 00:24:52,800 --> 00:24:58,880 attack is pretty costly, and you can make it more costly by having like a diverse ensemble of models 229 00:24:58,880 --> 00:25:03,680 for example, where an attack which works very well for one of the models doesn't work very well for 230 00:25:03,680 --> 00:25:07,760 the others because you are only gradient descending with respect to one of the models. 231 00:25:08,240 --> 00:25:13,520 With that being said, black box attacks are also apparently really good, which is super, 232 00:25:13,520 --> 00:25:19,520 super sad. The idea being that you can just train a similar model on effectively the same data set 233 00:25:19,520 --> 00:25:25,520 that the model you're attacking is, was trained on and get something like only like a half-redeemed 234 00:25:25,520 --> 00:25:30,400 was trained on and get something like only like a half reduction in your attack success. 235 00:25:31,120 --> 00:25:36,720 It's just tough, it's quite tough. I guess Florian's point of how many of these 236 00:25:36,720 --> 00:25:42,720 attackable models and how many of these attacks are possible in production level models, where like 237 00:25:42,720 --> 00:25:47,120 the model itself is hidden, you don't really know how like the data is cleaned, if the data is 238 00:25:47,120 --> 00:25:53,680 normalized, things like that is helpful, right? Security by obscurity. But at the end of the day, 239 00:25:53,680 --> 00:25:56,240 basically in theory, it's very hard to defend. 240 00:25:59,360 --> 00:26:02,000 I'm going to jump in a minute, but I just want to make one comment, which is that 241 00:26:02,000 --> 00:26:06,800 I think it's important to think about ZKML as just like one part of an overall crypto system. 242 00:26:06,800 --> 00:26:12,880 It's just like a single piece of a crypto system, you know, if you don't salt your passwords ever, 243 00:26:12,880 --> 00:26:16,640 right? Like you can make some mistake in the system that doesn't rate low people, 244 00:26:16,640 --> 00:26:20,480 but people do sort of obvious attacks. So I think whatever we design using ZKML, 245 00:26:20,480 --> 00:26:25,360 especially for security, we have to think about the context of the overall crypto system and how 246 00:26:25,360 --> 00:26:31,120 do we put the pieces together to make attacks like this difficult or impossible. Thank you for the 247 00:26:31,120 --> 00:26:39,600 discussion. Okay. 248 00:26:40,960 --> 00:26:48,720 Mic drop. Yeah, yeah, we are out of the time, but also I'm not sure about how much I have for 249 00:26:48,720 --> 00:26:56,000 recording, not on my laptop, but via Jitsi. So I think that for this call, we should be very 250 00:26:56,000 --> 00:27:03,440 satisfied of the overall result. I mean, at least I think that was so full of content. 251 00:27:04,640 --> 00:27:11,520 If you agree, but I think that you can ask also to the Telegram channel. I would love if we can 252 00:27:11,520 --> 00:27:18,240 do this frequently, maybe not every week, but maybe bi-weekly. So every two weeks. 253 00:27:19,280 --> 00:27:24,720 If this sounds good, we can plan for the, you know, for the middle of November or the second week 254 00:27:24,720 --> 00:27:32,480 of November. Do you like it? Sounds good to me. Yeah, I would suggest also starting with a round 255 00:27:32,480 --> 00:27:39,040 of intros so we know what everyone's doing and working on. Absolutely. Absolutely. I had some 256 00:27:39,040 --> 00:27:45,920 problem with the webcam, so I have to get back and forth, but yeah, next time we can dedicate half of 257 00:27:45,920 --> 00:27:54,400 the call, you know, to know better each other. Yeah, definitely. Please integrate the notes that 258 00:27:54,400 --> 00:28:04,400 DC Builders share in the chat with yours so we can publish in the description of the video that 259 00:28:04,400 --> 00:28:10,400 we are going to publish on YouTube and also in some repository for, you know, others to see. Okay. 260 00:28:10,400 --> 00:28:32,400 Cool. So let me say stop in the recording.
english