WEBVTT

1
00:00:00.152 --> 00:00:00.913
Hello, Internet.

2
00:00:00.933 --> 00:00:08.079
This is your captain speaking, Lester Nare, joined as always by my co-host and our resident PhD, Krishna Chowdhury.

3
00:00:08.440 --> 00:00:14.846
We have three great stories lined up for you this week, starting with a story out of the University of Toronto,

4
00:00:15.326 --> 00:00:20.568
which is around generative AI for cancer detection, which is published in Cell Genomics.

5
00:00:20.828 --> 00:00:21.629
This will be a good one.

6
00:00:22.009 --> 00:00:26.050
Our second story is a super story of a couple of institutions here involved.

7
00:00:26.451 --> 00:00:32.533
Science Corporation, Stanford University, UCSF, University of Pittsburgh, and University of Bonn.

8
00:00:32.553 --> 00:00:37.755
They've created a wireless retinal implant that is helping blind people see.

9
00:00:38.075 --> 00:00:40.356
This is out of the New England Journal of Medicine.

10
00:00:40.856 --> 00:00:44.378
And our final third story actually is two papers

11
00:00:45.058 --> 00:01:00.170
around the UFO subject, which is now known as Unidentified Anomalous Phenomenon, interestingly published in both Nature Scientific Reports and publications on the Astronomical Society of the Pacific.

12
00:01:00.770 --> 00:01:03.473
This is going to be really interesting, covering a lot of ground.

13
00:01:03.973 --> 00:01:06.135
As always, we're going to have a great time.

14
00:01:06.535 --> 00:01:08.977
This is From First Principles.

15
00:01:25.003 --> 00:01:25.443
My friend.

16
00:01:26.384 --> 00:01:26.925
How's it going?

17
00:01:27.525 --> 00:01:27.965
Quite well.

18
00:01:28.406 --> 00:01:28.846
Quite well.

19
00:01:29.206 --> 00:01:30.728
Back from our travels.

20
00:01:30.908 --> 00:01:31.668
Well, yeah.

21
00:01:31.688 --> 00:01:32.869
You know what happened.

22
00:01:33.290 --> 00:01:34.691
I mean, I do.

23
00:01:34.831 --> 00:01:35.031
I am.

24
00:01:36.712 --> 00:01:37.313
You know what happened.

25
00:01:38.472 --> 00:01:41.554
The Doyers won the World Series.

26
00:01:41.714 --> 00:01:44.596
Los Angeles brought home another World Championship.

27
00:01:44.716 --> 00:01:45.557
Back to back, baby.

28
00:01:45.617 --> 00:01:46.217
Back to back.

29
00:01:46.658 --> 00:01:47.318
As predicted.

30
00:01:47.398 --> 00:01:48.219
As predicted.

31
00:01:48.519 --> 00:01:50.440
We predicted the Chemistry Nobel Prize.

32
00:01:50.620 --> 00:01:50.841
Yes.

33
00:01:51.101 --> 00:01:52.722
And now we predicted the World Series.

34
00:01:52.782 --> 00:01:53.022
Yes.

35
00:01:53.322 --> 00:01:54.963
I mean, what can we not do?

36
00:01:55.044 --> 00:01:55.984
What can we not do?

37
00:01:56.124 --> 00:01:58.166
Yeah.

38
00:01:58.386 --> 00:02:02.028
LA is obviously just the greatest place to be right now.

39
00:02:02.048 --> 00:02:03.129
Yeah, yeah.

40
00:02:03.589 --> 00:02:05.191
I'm still re-watching.

41
00:02:06.836 --> 00:02:12.680
they're putting out like world series movies and I'm watching that

42
00:02:13.775 --> 00:02:15.216
A great, great Game 7.

43
00:02:15.356 --> 00:02:16.917
Yeah, it was an amazing Game 7.

44
00:02:17.097 --> 00:02:18.978
Amazing series, amazing Game 7.

45
00:02:19.098 --> 00:02:20.999
You know, shout out to the Jays.

46
00:02:21.439 --> 00:02:22.180
Put out a good fight.

47
00:02:22.340 --> 00:02:23.120
Yeah.

48
00:02:23.440 --> 00:02:31.605
And we're going to start off with a little bit of, you know, deference to Toronto, an incredible city.

49
00:02:31.965 --> 00:02:34.346
An incredible city, an incredible university.

50
00:02:34.406 --> 00:02:40.689
The University of Toronto is, you know, one of the great universities in the world of science and technology.

51
00:02:40.850 --> 00:02:41.090
Yes.

52
00:02:41.590 --> 00:02:43.491
They're kind of like the...

53
00:02:45.613 --> 00:03:13.626
neils boer institute when it comes to ai yes because of jeffrey hinton shout out to jeffrey hinton yeah i mean they've done a bunch of other things they discovered stem cells they discovered insulin so it's an incredible storied university and even though we defeated them even though we won even though we won let's let's just say we won let's let's just say we won and all these people being like you know i'm we got a comment on tiktok yeah saying that oh um have fun have

54
00:03:16.621 --> 00:03:20.682
And I finally feel like the Yankees did in the 2000s.

55
00:03:21.183 --> 00:03:25.024
And now I get it because I don't care either.

56
00:03:25.084 --> 00:03:25.564
I don't care.

57
00:03:25.584 --> 00:03:26.444
Yeah, whatever.

58
00:03:26.804 --> 00:03:32.746
But Toronto is a great city and the University of Toronto is a great university.

59
00:03:32.766 --> 00:03:33.146
Phenomenal.

60
00:03:33.166 --> 00:03:35.227
Phenomenal research institution.

61
00:03:35.247 --> 00:03:36.307
Phenomenal institution.

62
00:03:36.447 --> 00:03:37.808
And we're going to start off with –

63
00:03:38.668 --> 00:03:52.299
A story out of the University of Toronto, which is this new study about how AI-generated genomes promised to advance sort of precision cancer detection.

64
00:03:52.479 --> 00:03:52.740
Yes.

65
00:03:53.260 --> 00:03:54.961
And so that's sort of the story.

66
00:03:55.862 --> 00:03:58.404
And we cannot avoid...

67
00:04:00.165 --> 00:04:01.366
We cannot avoid AI.

68
00:04:01.526 --> 00:04:03.706
It's becoming much more prevalent.

69
00:04:03.746 --> 00:04:11.069
And it's interesting because a lot of times people think of generative AI as, oh, let me create an image about… Or chat GPT.

70
00:04:11.109 --> 00:04:11.449
Or chat GPT.

71
00:04:11.469 --> 00:04:13.590
The G in GPT is generative.

72
00:04:14.270 --> 00:04:16.210
And so a lot of people think of consumer use cases.

73
00:04:16.270 --> 00:04:24.093
But in sort of this early stage research arenas, there are some interesting things happening, this being one of them.

74
00:04:24.813 --> 00:04:32.180
So how exactly are the researchers at the University of Toronto using GenreVe AI for this detection process?

75
00:04:32.420 --> 00:04:32.620
Right.

76
00:04:32.981 --> 00:04:36.223
Well, the problem is precision oncology, okay?

77
00:04:36.784 --> 00:04:39.506
It needs a lot of large genomic data sets.

78
00:04:39.827 --> 00:04:40.007
Yep.

79
00:04:40.692 --> 00:04:47.017
But we also want to preserve the privacy of the individuals that are donating that data set, right?

80
00:04:47.397 --> 00:05:03.489
And so this creates a kind of bottleneck between the research and tool development for if we want to look at a genome, let's say of a patient, and try to figure out what is the specific cancer that this person has and how can we...

81
00:05:04.229 --> 00:05:05.210
Better mitigate that.

82
00:05:05.290 --> 00:05:07.974
What are the kinds of drugs that we need to actually do this?

83
00:05:08.355 --> 00:05:14.403
In order to actually train a model to do that job, we need a lot of data.

84
00:05:15.304 --> 00:05:19.630
But the data has to be, you know...

85
00:05:20.685 --> 00:05:23.889
without all of these privacy bottlenecks.

86
00:05:23.929 --> 00:05:27.273
The sort of personal identifiable information.

87
00:05:27.313 --> 00:05:28.154
Exactly, yeah.

88
00:05:28.234 --> 00:05:29.656
And that's what's tough, okay?

89
00:05:30.036 --> 00:05:34.481
And this Oncogan, which is a generative AI pipeline, it was published in...

90
00:05:36.680 --> 00:05:46.589
I believe cell genomics, cell genomics, Oncogan, it creates a highly realistic synthetic cancer genome just completely within the computer.

91
00:05:47.049 --> 00:05:47.310
Okay.

92
00:05:47.550 --> 00:05:49.472
And it bypasses these privacy issues.

93
00:05:49.512 --> 00:05:50.593
That's actually really interesting.

94
00:05:50.613 --> 00:05:50.733
Okay.

95
00:05:50.753 --> 00:05:50.973
Right.

96
00:05:51.433 --> 00:06:01.642
And the key feature is it mimics all of these computational, complex, mutational landscapes that are characteristic of cancers.

97
00:06:02.506 --> 00:06:05.371
But it doesn't tie that data with the patient.

98
00:06:05.391 --> 00:06:06.452
With an individual.

99
00:06:07.227 --> 00:06:22.538
Yeah, and the goal is basically to democratize access to this high-fidelity genomic data, accelerate cancer diagnosis, treatment, prevention tools, all of the stuff that we want without actually compromising patient confidentiality.

100
00:06:23.018 --> 00:06:35.687
The idea being if you can provide this sort of substrate of the data set, a variety of research teams and institutions can do work towards solving the problem without running, you know,

101
00:06:36.494 --> 00:06:38.135
Into all of the regulations.

102
00:06:38.155 --> 00:06:38.916
Regulation issues.

103
00:06:38.936 --> 00:06:39.996
Which are there for a good reason.

104
00:06:40.016 --> 00:06:41.577
Which are there for a good reason.

105
00:06:41.737 --> 00:06:45.299
So let's go through what precision oncology means.

106
00:06:45.520 --> 00:06:45.780
Okay.

107
00:06:46.000 --> 00:06:50.002
Why it's actually hard to actually simulate these genomes.

108
00:06:50.102 --> 00:06:50.303
Yeah.

109
00:06:50.483 --> 00:06:52.924
And then what this current paper is doing.

110
00:06:53.044 --> 00:06:53.284
Okay.

111
00:06:53.304 --> 00:06:53.565
Okay.

112
00:06:53.625 --> 00:06:55.226
What this current paper is announcing.

113
00:06:55.546 --> 00:07:01.950
So the vision of precision oncology is we want to shift from things like chemotherapy and radiation, which is...

114
00:07:02.590 --> 00:07:12.833
kind of just a general sort of shotgun approach to this tailored treatment based on a patient's tumor, the molecular and genomic alterations in that tumor.

115
00:07:12.913 --> 00:07:13.173
Okay.

116
00:07:13.453 --> 00:07:14.873
And we want to make it precision.

117
00:07:14.913 --> 00:07:16.654
That's, that's where precision comes from.

118
00:07:16.954 --> 00:07:17.134
Right.

119
00:07:17.394 --> 00:07:21.695
And so what we want to do is identify mutations and then administer targeted drugs.

120
00:07:22.235 --> 00:07:27.377
There's a data bottleneck where we have this paradox between sharing and privacy, right?

121
00:07:27.417 --> 00:07:28.197
Because privacy,

122
00:07:28.760 --> 00:07:41.246
The success of precision oncology and an AI to precisely identify what type of oncology is going on depends on a vast, diverse data set, right?

123
00:07:41.506 --> 00:07:47.029
But every single genomic identifier is something that is unique to every individual, right?

124
00:07:47.049 --> 00:07:48.550
You've got 3 billion base pairs.

125
00:07:48.590 --> 00:07:52.251
And if I know your 3 billion base pairs, I can figure out who you are, right?

126
00:07:52.792 --> 00:07:53.032
So...

127
00:07:53.932 --> 00:07:56.033
re-identification is really possible.

128
00:07:56.113 --> 00:08:17.606
Like even if you de-identify data, like you took some patients' genomic data and you took out some parts, it would actually be pretty easy to look up who you are based on genealogical websites, public records, the whole like 23andMe thing, Ancestry.com, like that whole thing, right?

129
00:08:17.926 --> 00:08:19.387
So there's privacy risks,

130
00:08:20.167 --> 00:08:24.571
And that can reveal dispositions to disease, potential discrimination.

131
00:08:24.891 --> 00:08:30.576
And you've got stuff like HIPAA that actually controls and creates barriers for that kind of data sharing.

132
00:08:30.816 --> 00:08:32.758
It's a very good thing that we're doing that, right?

133
00:08:32.778 --> 00:08:38.623
Because then without it, you can have potential discrimination on life insurance, long-term care.

134
00:08:38.643 --> 00:08:38.743
Yeah.

135
00:08:39.023 --> 00:08:39.243
Right.

136
00:08:39.324 --> 00:08:49.477
I was going to say that the big issue is you don't want insurance providers to be able to basically say, oh, because we know that you have – this is the whole Obamacare preexisting conditions issue.

137
00:08:49.637 --> 00:08:49.917
Yeah.

138
00:08:49.977 --> 00:08:57.167
If we know who you are and what your preexisting conditions are, we're going to charge you more because you're a higher risk and liability for us as the insurance provider.

139
00:08:57.307 --> 00:08:57.747
Exactly.

140
00:08:57.787 --> 00:08:57.967
Yeah.

141
00:08:57.987 --> 00:08:58.668
You don't want that.

142
00:08:58.968 --> 00:09:01.069
That's just blatant nonsense.

143
00:09:01.089 --> 00:09:01.429
Yeah.

144
00:09:01.509 --> 00:09:01.709
Right.

145
00:09:02.430 --> 00:09:13.256
The other thing actually that it's tackling is if you have an AI generated tool, whenever we make AI, you want to have a ground truth data set that tells you what is right or wrong.

146
00:09:13.636 --> 00:09:13.856
Right.

147
00:09:14.216 --> 00:09:17.618
So that you can compare one AI model from another.

148
00:09:18.179 --> 00:09:24.122
Well, with with cancer data sets, it's really hard to do that kind of benchmarking because.

149
00:09:25.761 --> 00:09:29.402
The real patient samples have sequencing errors.

150
00:09:29.782 --> 00:09:32.423
You don't actually know what the tumor complexity is.

151
00:09:32.723 --> 00:09:34.464
There's no actual ground truth.

152
00:09:34.864 --> 00:09:42.787
So if we were to generate data sets that have a ground truth, then we can easily benchmark all of these other AI models.

153
00:09:43.167 --> 00:09:46.268
So when someone says, hey, I've got an AI model that I think does better—

154
00:09:46.848 --> 00:09:49.029
We've actually got a data set where we can test that.

155
00:09:49.109 --> 00:09:49.370
Yeah.

156
00:09:49.650 --> 00:09:49.850
Right?

157
00:09:49.910 --> 00:09:50.430
Yeah.

158
00:09:50.550 --> 00:09:53.652
So that's a separate issue from the privacy thing.

159
00:09:53.993 --> 00:09:57.015
But this thing is actually tackling both of them at the same time.

160
00:09:57.075 --> 00:09:57.395
Interesting.

161
00:09:57.415 --> 00:10:00.077
And obviously there's a huge economic and human imperative, right?

162
00:10:00.417 --> 00:10:07.622
Like there's a global cancer burden, 20 million new cases every year, 9.7 million deaths every year.

163
00:10:07.942 --> 00:10:09.383
There's a huge economic toll.

164
00:10:10.324 --> 00:10:15.029
Drug development is long and expensive, takes hundreds of millions of dollars.

165
00:10:15.529 --> 00:10:21.235
And if we have this kind of data set, we can train models that do it way better, right?

166
00:10:21.295 --> 00:10:24.859
We can limit all of that and democratize.

167
00:10:25.658 --> 00:10:32.562
This is actually totally an organic anecdote, but literally I was in a work meeting earlier today.

168
00:10:33.183 --> 00:10:41.007
And unfortunately, one of my coworkers, you know, has a partner that was just diagnosed with stage four.

169
00:10:41.628 --> 00:10:44.670
And like, you know, I mean, it happens every day.

170
00:10:44.690 --> 00:10:45.530
Yeah.

171
00:10:45.650 --> 00:10:46.311
It is present.

172
00:10:47.111 --> 00:10:51.213
Personally, for millions of people, one, two degrees of separation for so many people.

173
00:10:51.233 --> 00:10:52.293
Yeah, everyone knows someone.

174
00:10:52.393 --> 00:10:53.314
And literally today.

175
00:10:53.454 --> 00:11:02.778
I mean, so this is clearly, obviously, a highly important and prevalent issue.

176
00:11:02.998 --> 00:11:03.798
Yes, exactly.

177
00:11:04.078 --> 00:11:10.061
And so let's talk about like cancer itself and why it's even like hard to do this.

178
00:11:10.621 --> 00:11:15.202
It's hard to like make these genomic data sets that are synthetic the way that we want.

179
00:11:15.562 --> 00:11:15.803
Okay.

180
00:11:16.143 --> 00:11:17.683
Cancer is a disease of the genome.

181
00:11:17.963 --> 00:11:19.244
It's caused by DNA errors.

182
00:11:19.364 --> 00:11:21.184
At the end of the day, that's what's happening.

183
00:11:21.305 --> 00:11:21.525
Okay.

184
00:11:21.985 --> 00:11:26.746
There's 3 billion DNA letters, ATGC, in our genome.

185
00:11:27.467 --> 00:11:27.587
And

186
00:11:28.367 --> 00:11:31.970
We're going to use this analogy of the genome being a library of cookbooks.

187
00:11:32.350 --> 00:11:43.099
Every single chromosome is a single cookbook that, let's say, one cookbook is for Indian food, one cookbook is for desserts, and so on and so forth, right?

188
00:11:43.539 --> 00:11:46.322
So the genome is a library of these cookbooks.

189
00:11:46.722 --> 00:11:51.686
Each cookbook is a chromosome, and then each recipe is a gene inside that cookbook.

190
00:11:51.866 --> 00:11:52.147
Okay.

191
00:11:52.347 --> 00:11:52.627
Okay?

192
00:11:53.974 --> 00:11:58.076
If we have errors in critical recipes, that's going to lead to cancer.

193
00:11:59.236 --> 00:12:06.240
And there's really two types of big mutations that happen, these typos that happen in our recipe.

194
00:12:06.560 --> 00:12:06.800
Okay?

195
00:12:07.060 --> 00:12:17.605
There's germline mutations, which is basically when a mutation is, right, when a letter gets substituted for another letter or some mistake happens in the genomic process.

196
00:12:18.405 --> 00:12:23.731
reading of the thing, or I delete a bunch of letters, so on and so forth, right?

197
00:12:24.151 --> 00:12:26.374
There's two types of mutations that happen.

198
00:12:26.394 --> 00:12:32.440
There's the germline mutation, that's something that's inherited from your parents, and that's present in every single one of your cells, right?

199
00:12:32.500 --> 00:12:39.107
Because that's the, you became, you came from a zygote, which is a single cell, and so whatever mutation was in that zygote,

200
00:12:39.988 --> 00:12:41.269
That's all of you, right?

201
00:12:41.590 --> 00:12:43.291
Then there's somatic mutations.

202
00:12:43.551 --> 00:12:45.273
These are acquired during your lifetime.

203
00:12:45.693 --> 00:12:47.114
It happens within a single cell.

204
00:12:47.434 --> 00:12:52.238
And then the progeny of that single cell inherits that somatic mutation, right?

205
00:12:52.739 --> 00:12:57.523
Cancer is actually primarily driven by accumulated somatic mutations.

206
00:12:57.543 --> 00:12:57.803
Okay.

207
00:12:57.843 --> 00:12:58.103
Yeah.

208
00:13:00.285 --> 00:13:04.807
There are obviously cases where germline mutations create the cancer off the bat.

209
00:13:05.048 --> 00:13:05.228
Yes.

210
00:13:05.248 --> 00:13:18.195
But usually it happens with age, and there's these accumulated somatic mutations that cause your cells to do weird things, become tumorous, become carcinogenic, and then create the cancer, right?

211
00:13:18.215 --> 00:13:18.996
That makes sense.

212
00:13:19.546 --> 00:13:26.670
One key aspect is Oncogan, which is this generative AI, it's trained on and it generates only somatic mutations.

213
00:13:27.190 --> 00:13:30.952
It doesn't get trained on the germline info.

214
00:13:31.072 --> 00:13:35.714
So it's already getting rid of that sort of privacy issue.

215
00:13:36.534 --> 00:13:43.378
Which is important to understand that there's two entry points, one of which has less of a privacy concern by default.

216
00:13:44.418 --> 00:13:46.939
And that's where this Oncogan is focused on.

217
00:13:46.959 --> 00:13:48.160
Exactly, exactly, yeah.

218
00:13:48.440 --> 00:13:51.701
Now, when it comes to the architects of cancers, there's two types of cancers.

219
00:13:51.741 --> 00:13:53.462
There's a driver and a passenger.

220
00:13:53.762 --> 00:13:59.484
So the driver mutation, these are critical alterations that give a growth advantage to some kind of cell.

221
00:13:59.804 --> 00:14:00.064
Okay?

222
00:14:00.424 --> 00:14:01.885
It can either be something like...

223
00:14:02.730 --> 00:14:12.521
you turn on an oncogene, which is something that turns on the cancer part of the genome, or it turns off a tumor repressor, right?

224
00:14:12.802 --> 00:14:13.803
A tumor suppressor.

225
00:14:14.123 --> 00:14:20.991
So then if the cell wants to go become tumorous, this gene isn't being active telling it, no, you don't want to do that, right?

226
00:14:21.011 --> 00:14:21.431
No, not today.

227
00:14:23.354 --> 00:14:25.337
So these are the driver mutations.

228
00:14:25.698 --> 00:14:28.783
Then there's passenger mutations, which are kind of like hitchhikers.

229
00:14:28.823 --> 00:14:29.984
They accumulate by chance.

230
00:14:30.365 --> 00:14:31.567
There's no growth advantage.

231
00:14:32.028 --> 00:14:37.096
It's kind of like just there, but it doesn't really affect the...

232
00:14:37.516 --> 00:14:42.141
the cellular machinery in the way that it like turns it on and becomes cancerous, right?

233
00:14:42.461 --> 00:14:49.589
Tumors have very few driver mutations amidst thousands of passenger mutations.

234
00:14:49.789 --> 00:14:50.049
Okay.

235
00:14:50.269 --> 00:14:57.317
So there's this distribution where there's a few that are targeted for these oncogenes or these tumor suppressor genes that,

236
00:14:58.054 --> 00:14:59.275
And then there's a bunch of noise.

237
00:14:59.615 --> 00:14:59.935
Yes.

238
00:15:00.035 --> 00:15:00.255
Right?

239
00:15:00.796 --> 00:15:10.062
And any kind of generative model like Oncogan needs to model that discrepancy where there's a few of the stuff that matters and there's a lot of noise.

240
00:15:10.442 --> 00:15:10.602
Right?

241
00:15:10.902 --> 00:15:12.543
And Oncogan does exactly that.

242
00:15:12.723 --> 00:15:12.923
Right?

243
00:15:13.044 --> 00:15:15.365
It replicates this tumor architecture perfectly.

244
00:15:15.887 --> 00:15:20.812
by making a bunch of random passengers and not that many driver mutations.

245
00:15:21.373 --> 00:15:21.613
Okay?

246
00:15:22.334 --> 00:15:28.301
And the final thing I want to touch on is these mutations have fingerprints when it comes to cancer.

247
00:15:28.581 --> 00:15:28.841
Okay?

248
00:15:29.362 --> 00:15:33.126
There's characteristic patterns of mutations that...

249
00:15:34.323 --> 00:15:38.807
give certain processes a leg up that becomes cancerous.

250
00:15:39.087 --> 00:15:40.729
For example, like UV light, right?

251
00:15:40.869 --> 00:15:43.912
UV light causes a C to T mutation.

252
00:15:44.312 --> 00:15:53.921
Basically you get this dimerization where like one leg of your DNA, you know, your DNA is a twisted ladder, but one part of your ladder is going to like staple itself.

253
00:15:54.201 --> 00:15:54.481
Right?

254
00:15:54.721 --> 00:15:57.104
Like one leg of your ladder, it's got two legs.

255
00:15:57.825 --> 00:16:02.210
One leg of your ladder is going to staple itself onto itself and become a dimer.

256
00:16:02.711 --> 00:16:06.035
And then that's going to cause melanoma, for example, right?

257
00:16:06.415 --> 00:16:12.302
Tobacco smoke is linked to another single base substitution that'll cause lung cancer.

258
00:16:13.275 --> 00:16:22.765
So Oncogan, which is this thing that people have made, it accurately reproduces that tissue-specific signature, right?

259
00:16:22.785 --> 00:16:29.212
Where it's like the lung cancer is going to do this and the skin cancer is going to do that, so on and so forth.

260
00:16:29.272 --> 00:16:32.075
It's actually learned how to do all of that stuff.

261
00:16:32.115 --> 00:16:32.916
That's fascinating.

262
00:16:33.336 --> 00:16:37.698
And then the other thing, we're not done with all of the stuff that can cause cancer.

263
00:16:37.998 --> 00:16:40.720
There's also something called large-scale vandalism, right?

264
00:16:40.760 --> 00:16:43.861
There's copy number alterations, CNAs.

265
00:16:44.782 --> 00:16:47.283
This means you've taken a whole chunk of your chromosome

266
00:16:48.294 --> 00:16:50.016
And you've duplicated it.

267
00:16:50.456 --> 00:16:50.917
Okay.

268
00:16:51.117 --> 00:16:51.357
Okay.

269
00:16:51.637 --> 00:16:56.542
So now there's a duplication or even a deletion sometimes of large chromosomal segments.

270
00:16:56.882 --> 00:17:03.428
So you've got large parts, like large chapters of your cookbook that are now been copied or completely deleted.

271
00:17:03.749 --> 00:17:08.973
You've also got structural variants where you take a big chunk of your DNA and you flip it.

272
00:17:09.394 --> 00:17:10.415
And now it's reversed.

273
00:17:11.497 --> 00:17:13.497
Now, it's just all sorts of chaos, right?

274
00:17:13.517 --> 00:17:14.118
Right, right, right.

275
00:17:14.158 --> 00:17:19.899
Like, there's so many different things that can happen with cancer genomes, right?

276
00:17:19.939 --> 00:17:21.559
That are the driver and the cause of it.

277
00:17:21.699 --> 00:17:22.219
Right.

278
00:17:22.399 --> 00:17:35.402
And the real sort of challenge is how do we make a single architecture, a single framework that captures all of these different myriad effects, right, into a single pipeline?

279
00:17:35.422 --> 00:17:35.602
Mm-hmm.

280
00:17:36.733 --> 00:17:41.523
that then can be used to train other AI, to train other tools, right?

281
00:17:41.864 --> 00:17:43.166
And that's what this paper is doing.

282
00:17:43.246 --> 00:17:47.876
So they use something called generative AI.

283
00:17:49.170 --> 00:17:50.391
We've seen this a lot, right?

284
00:17:50.451 --> 00:17:52.292
Like a cat on a thing.

285
00:17:52.712 --> 00:17:56.094
This one, this particular architecture is called a GAN.

286
00:17:56.194 --> 00:17:58.675
It's a generative adversarial network, okay?

287
00:17:59.056 --> 00:18:00.817
There's two competing neural networks.

288
00:18:00.897 --> 00:18:06.280
This is a very interesting way to actually train AI and to architecture it up.

289
00:18:06.660 --> 00:18:07.420
I think it's very cool.

290
00:18:07.440 --> 00:18:09.902
It's been used actually for image generation as well.

291
00:18:10.042 --> 00:18:10.242
Yep.

292
00:18:10.782 --> 00:18:15.744
But this time what they're doing is they're using it to generate genomic data.

293
00:18:16.064 --> 00:18:24.007
It's interesting because I've been watching this space for a while and I've seen GANs being used in the image generation context and...

294
00:18:25.335 --> 00:18:32.744
Now mapping it to this problem set is an interesting use of the architecture of generative adversarial networks.

295
00:18:32.764 --> 00:18:33.765
That's interesting.

296
00:18:34.286 --> 00:18:40.654
And the basic idea behind generative adversarial networks is you've got two networks, okay?

297
00:18:40.974 --> 00:18:42.556
You've got one that's the forger.

298
00:18:43.357 --> 00:18:45.298
And one that's the discriminator.

299
00:18:45.838 --> 00:18:46.119
Okay?

300
00:18:46.859 --> 00:18:54.423
The forger or the generator, this is the guy who's going to create forgeries from random noise and try to attempt realism.

301
00:18:54.843 --> 00:19:05.629
And then there's the discriminator, which is like the judge who's trying to tell whether the thing that was generated is a real thing or a fake thing.

302
00:19:06.149 --> 00:19:10.452
So both of these guys have the training set, which is all the stuff that is actually real.

303
00:19:10.792 --> 00:19:14.434
And the generator is creating fake versions of that.

304
00:19:14.614 --> 00:19:14.755
Yep.

305
00:19:15.275 --> 00:19:20.198
And over many, many iterations, both of these guys are going to get really good at their jobs.

306
00:19:20.598 --> 00:19:23.700
The generator is going to get really good at creating fake data.

307
00:19:24.601 --> 00:19:25.301
And the...

308
00:19:26.642 --> 00:19:30.686
discriminator is going to get really good at telling whether it's fake or not.

309
00:19:31.287 --> 00:19:41.898
And it's this competing adversarial, that's where we get adversarial from, this competing effect that actually causes the generator to get really, really good at its job.

310
00:19:42.439 --> 00:19:47.704
So there's a substrate of truth that these two processes have access to.

311
00:19:48.645 --> 00:19:53.088
One's job is to create synthetic data based on that substrate.

312
00:19:53.168 --> 00:19:53.408
Yeah.

313
00:19:53.568 --> 00:19:56.469
The other is to say whether you're fake or not.

314
00:19:57.890 --> 00:20:05.774
And because there's this sort of iterative back and forth between these two, you get to a place where –

315
00:20:06.755 --> 00:20:16.880
you know, the generator has now generated something that the discriminator views as being real because it's gone through that back and forth process.

316
00:20:16.900 --> 00:20:17.240
Exactly.

317
00:20:17.280 --> 00:20:21.522
And now it's like really good at making real looking data.

318
00:20:21.642 --> 00:20:25.024
Which goes back to our privacy issue and some of the other things we just talked about.

319
00:20:25.044 --> 00:20:25.204
Exactly.

320
00:20:25.224 --> 00:20:25.424
Yeah.

321
00:20:25.564 --> 00:20:25.744
Yeah.

322
00:20:26.084 --> 00:20:31.807
And so this cycle repeats until the generator produces stuff that's indistinguishable from the real data.

323
00:20:31.827 --> 00:20:31.967
Yeah.

324
00:20:32.727 --> 00:20:36.772
And this can be used for tabular data, which is a lot of this genomic data, right?

325
00:20:36.832 --> 00:20:38.474
It's not like images and things like that.

326
00:20:38.494 --> 00:20:41.717
It's like tables of like how much of this is there, how much of this is there.

327
00:20:42.398 --> 00:20:47.143
Oncogan uses something called CTABGAN+, which is just...

328
00:20:48.044 --> 00:20:50.466
specific architecture that's used for tabular data.

329
00:20:50.666 --> 00:20:50.926
Got it.

330
00:20:51.227 --> 00:20:54.569
And it handles mixed data types, imbalanced distributions.

331
00:20:54.909 --> 00:20:57.411
So like the distribution is not like completely uniform.

332
00:20:57.731 --> 00:20:59.532
You have some stuff happening all the time.

333
00:20:59.813 --> 00:21:01.254
Some mutations happen all the time.

334
00:21:01.674 --> 00:21:03.755
Other mutations are rare, so on and so forth.

335
00:21:03.775 --> 00:21:04.856
So this takes care of that.

336
00:21:04.876 --> 00:21:07.758
It handles the complexity of this kind of data set.

337
00:21:07.898 --> 00:21:08.379
Exactly.

338
00:21:08.499 --> 00:21:08.939
Exactly.

339
00:21:09.319 --> 00:21:12.782
Now, the other thing that they're using is something called variational auto encoders.

340
00:21:13.282 --> 00:21:14.123
These are,

341
00:21:16.062 --> 00:21:17.543
simpler than GANs.

342
00:21:17.903 --> 00:21:27.188
And what they do is they basically take whatever input you have, they try to squish it down into a tiny little latent space, right?

343
00:21:27.769 --> 00:21:33.992
And that latent space has fewer dimensions than the original data.

344
00:21:34.492 --> 00:21:42.477
And then from that latent space, we have a decoder that tries to recreate the...

345
00:21:42.881 --> 00:21:45.343
what was originally compressed.

346
00:21:45.443 --> 00:21:45.883
Yeah.

347
00:21:45.943 --> 00:21:46.223
Okay?

348
00:21:46.323 --> 00:21:46.563
Yeah.

349
00:21:46.683 --> 00:21:48.325
So this is used a lot.

350
00:21:48.405 --> 00:21:51.166
This is used actually a lot in like image generation.

351
00:21:51.847 --> 00:22:00.412
When we have, let's say a 256 by 256 image, that has 256 times 256 dimensions, right?

352
00:22:00.452 --> 00:22:02.494
Because each pixel has a value.

353
00:22:02.774 --> 00:22:09.578
And you can imagine if there's only three pixels, you can imagine putting that on an XYZ plane, right?

354
00:22:14.532 --> 00:22:17.314
Like the first pixel is how far it is in the X direction.

355
00:22:17.814 --> 00:22:21.537
The second pixel is how far it is in the Y direction, and so on and so forth.

356
00:22:21.797 --> 00:22:25.700
Well, now you have 256 times 256 dimensions.

357
00:22:26.320 --> 00:22:28.842
Every single image is highly dimensional.

358
00:22:29.302 --> 00:22:36.847
But if I were to take a bunch of photos of just a bunch of faces, right, the essence of the face is actually...

359
00:22:38.243 --> 00:22:40.104
not that highly dimensional.

360
00:22:40.384 --> 00:22:41.845
There's always a nose in the middle.

361
00:22:42.105 --> 00:22:42.965
There's always a mouth.

362
00:22:42.985 --> 00:22:45.306
There's always eyes, right?

363
00:22:45.446 --> 00:22:49.608
Maybe one of the directions could be how dark is the face, right?

364
00:22:49.628 --> 00:22:51.569
For us, it would be like maxed out.

365
00:22:52.429 --> 00:22:53.049
Very dark.

366
00:22:53.069 --> 00:22:53.869
Yeah, very dark.

367
00:22:54.070 --> 00:23:00.372
Like there would be one dimension that would tell the variational autoencoder, this is a very dark face, right?

368
00:23:00.592 --> 00:23:06.555
Then there would be one dimension that would tell how far up in the nose is the nose and how wide is the mouth and things like that.

369
00:23:06.935 --> 00:23:07.135
But...

370
00:23:08.090 --> 00:23:16.072
The fact that there is a mouth is not something that needs to be encoded in the latent space because it's always there, right?

371
00:23:16.092 --> 00:23:16.432
That makes sense.

372
00:23:16.592 --> 00:23:22.354
The stuff that's varying is the stuff that's going to be encoded in that compressed middle part, right?

373
00:23:22.434 --> 00:23:28.556
And then you have a decoder that takes that middle part that says, okay, how much is the, how much is, where is this and where is that?

374
00:23:28.596 --> 00:23:31.736
And puts it, puts it into the actual image that it reconstructs.

375
00:23:32.117 --> 00:23:34.197
And this is something that we're seeing in MNIST.

376
00:23:34.477 --> 00:23:37.398
For example, the MNIST dataset is a bunch of handwritten digits.

377
00:23:38.339 --> 00:23:47.718
The key part about variational autoencoders is that it's a continuous latent space, meaning that all of my real data points are somewhere in my...

378
00:23:48.664 --> 00:23:54.647
latent space, but I can pick any point in between and that'll result in something real.

379
00:23:55.168 --> 00:23:55.508
Okay.

380
00:23:55.588 --> 00:23:56.829
It's continuous.

381
00:23:56.969 --> 00:24:01.411
It's not discrete like this and this, and there's nothing in between, right?

382
00:24:01.731 --> 00:24:04.113
There is actually meaning in between.

383
00:24:04.333 --> 00:24:12.498
And so you can see in that latent space of the MNIST, it's going from the thing that looks like a nine, which is over here.

384
00:24:12.618 --> 00:24:12.818
Yeah.

385
00:24:12.838 --> 00:24:12.938
Yeah.

386
00:24:13.358 --> 00:24:16.260
to a three, to a six, to a zero, right?

387
00:24:16.340 --> 00:24:19.062
And it's continuously deforming.

388
00:24:19.122 --> 00:24:25.127
And all they're doing is reconstructing what happens if I move on that latent space, like as a walk.

389
00:24:25.567 --> 00:24:28.870
And I see, what is the decoder reconstructing?

390
00:24:28.890 --> 00:24:30.071
Does that sort of make sense?

391
00:24:30.171 --> 00:24:31.071
No, 100%.

392
00:24:32.092 --> 00:24:37.997
It's like, it's taking all of the information of the stuff that it's trained on and saying, what is the essence?

393
00:24:38.197 --> 00:24:38.417
Yes.

394
00:24:38.637 --> 00:24:39.158
Of the thing.

395
00:24:39.238 --> 00:24:39.438
Yes.

396
00:24:39.618 --> 00:24:42.581
I'm going to put that into a landscape that I can traverse.

397
00:24:42.621 --> 00:24:43.142
You can traverse.

398
00:24:43.162 --> 00:24:43.843
Yeah, exactly.

399
00:24:44.183 --> 00:24:49.068
And then as I move through that landscape, it's going to create meaning with my decoder.

400
00:24:49.288 --> 00:24:55.575
Because there's different points on that landscape, for example, in the numbers that we find meaning in, which is like the nine...

401
00:24:56.035 --> 00:24:56.755
The three.

402
00:24:57.135 --> 00:24:57.796
The zero.

403
00:24:58.276 --> 00:25:01.917
But there is this transformation.

404
00:25:02.057 --> 00:25:02.797
Continuous.

405
00:25:02.997 --> 00:25:09.579
Continuous in between there where you can say, oh, this is like a nine, three-ish thingy.

406
00:25:10.479 --> 00:25:10.739
Yeah.

407
00:25:10.899 --> 00:25:12.259
On my way from nine to three.

408
00:25:12.299 --> 00:25:13.360
Exactly.

409
00:25:13.400 --> 00:25:14.480
Yeah, exactly.

410
00:25:15.060 --> 00:25:15.520
Exactly.

411
00:25:15.540 --> 00:25:16.460
Yeah, yeah.

412
00:25:16.601 --> 00:25:18.161
So they're using both.

413
00:25:18.341 --> 00:25:21.762
And the key to using both is that you've got, again...

414
00:25:22.800 --> 00:25:31.823
that's used for complex interdependent features like mutation counts, driver co-occurrence, that's there due to high fidelity matching.

415
00:25:32.203 --> 00:25:37.965
And then you have this variational autoencoder that's there for continuous variables, like genomic position.

416
00:25:38.265 --> 00:25:40.385
The position is a continuous thing, right?

417
00:25:40.825 --> 00:25:43.166
But the mutation count, that's a discrete thing.

418
00:25:43.186 --> 00:25:43.826
That's a discrete thing.

419
00:25:43.866 --> 00:25:44.046
Right?

420
00:25:44.347 --> 00:25:48.968
So you're combining both into this hybrid approach, right?

421
00:25:49.228 --> 00:25:52.555
To create this realism in the data set.

422
00:25:52.595 --> 00:25:55.380
That's actually really, really interesting.

423
00:25:55.420 --> 00:25:59.067
They're not limiting themselves to one type of architecture.

424
00:25:59.087 --> 00:25:59.528
Right, right.

425
00:25:59.548 --> 00:26:00.129
They're really...

426
00:26:01.315 --> 00:26:02.996
Harnessing the true power.

427
00:26:03.176 --> 00:26:10.339
It's like there's this blended architecture which is using the value of both of these modalities and taking the best of both.

428
00:26:10.539 --> 00:26:10.799
Yeah.

429
00:26:11.060 --> 00:26:19.504
And sort of using them in the context that they're good at in parallel or together.

430
00:26:19.544 --> 00:26:20.684
Yeah, exactly.

431
00:26:21.064 --> 00:26:21.424
Exactly.

432
00:26:21.444 --> 00:26:24.966
Because the genome is both continuous and discrete.

433
00:26:25.046 --> 00:26:26.147
And so now we're...

434
00:26:27.488 --> 00:26:31.953
We're getting like two soldiers here that are doing the work, right?

435
00:26:32.393 --> 00:26:40.302
In order to train Oncogan, they used the training data from Pan Cancer Analysis of Whole Genomes, PCAWG.

436
00:26:40.682 --> 00:26:42.164
It's a data set that has 2,658 genes.

437
00:26:43.986 --> 00:26:46.929
whole cancer genomes, like entire genomes, okay?

438
00:26:47.249 --> 00:26:49.011
All three billion base pairs.

439
00:26:49.031 --> 00:26:49.291
Massive.

440
00:26:49.471 --> 00:26:50.032
Massive.

441
00:26:50.272 --> 00:26:50.692
Massive.

442
00:26:50.912 --> 00:26:56.418
Yeah, and this data set is treated like sensitive information because it should be, right?

443
00:26:56.478 --> 00:27:01.202
Because this has actual markers that can identify patients and things like that.

444
00:27:01.622 --> 00:27:09.109
But using that, they've now created something that will generate data that doesn't have those...

445
00:27:10.129 --> 00:27:21.958
They've created a derivative product from source data that is personal identifiable information that now has that degree of separation to not run afoul of the privacy.

446
00:27:21.978 --> 00:27:22.298
Exactly.

447
00:27:22.338 --> 00:27:23.519
Yeah, yeah, exactly.

448
00:27:23.599 --> 00:27:28.042
And from what I counted in the paper, there's five GAN models and one TVA model.

449
00:27:28.362 --> 00:27:35.307
So there's five GAN generative adversarial network models and then one variational autoencoder model.

450
00:27:36.303 --> 00:27:41.451
all of these guys are working together to create this giant genome, right?

451
00:27:41.571 --> 00:27:42.733
Multiple giant genomes.

452
00:27:43.174 --> 00:27:45.858
And obviously the question is to ask is like, you know,

453
00:27:46.697 --> 00:27:49.018
Okay, how real are you making this, right?

454
00:27:49.158 --> 00:27:51.479
The photo that you're making, could it fool me?

455
00:27:52.060 --> 00:27:57.042
This becomes the, with photos, it's easy because we sort of have the eye test, right?

456
00:27:57.062 --> 00:28:03.625
This is the whole uncanny valley concept, which is like, okay, I can look at something and be like, oh, the fingers aren't right.

457
00:28:03.665 --> 00:28:06.026
I mean, now, dude, I'll be honest, dude.

458
00:28:06.046 --> 00:28:09.628
There's like videos on Instagram that I get fooled.

459
00:28:10.608 --> 00:28:13.451
I do a double take and I'm like, oh, this is definitely.

460
00:28:13.611 --> 00:28:18.855
There's a Neil deGrasse Tyson just did a video of himself talking about how earth is flat.

461
00:28:19.176 --> 00:28:21.397
And he held up an iPad to the camera.

462
00:28:21.417 --> 00:28:22.198
You didn't know.

463
00:28:22.258 --> 00:28:25.481
The video starts with just Neil deGrasse Tyson talking about how the earth is flat.

464
00:28:25.801 --> 00:28:29.544
And then he just moves the iPad and he's like, that wasn't me.

465
00:28:30.225 --> 00:28:30.505
And it's.

466
00:28:31.566 --> 00:28:32.227
It's like, Oh my God.

467
00:28:32.247 --> 00:28:32.907
And it's the same.

468
00:28:32.967 --> 00:28:34.388
It's like him in the same space.

469
00:28:34.508 --> 00:28:35.129
Oh my God.

470
00:28:35.149 --> 00:28:37.210
The same outfit and the same voice.

471
00:28:37.851 --> 00:28:40.353
And it's effectively indistinguishable.

472
00:28:40.838 --> 00:28:43.479
Yeah, dude, it's getting really good, right?

473
00:28:43.499 --> 00:28:44.000
It's quite good.

474
00:28:44.020 --> 00:28:46.921
So at least we're harnessing it for something useful.

475
00:28:46.941 --> 00:28:48.782
Yeah.

476
00:28:49.022 --> 00:28:58.387
So one of the things they looked at to see, okay, is the stuff that's being output by this neural network, is it something that can be real, right?

477
00:28:58.667 --> 00:29:00.548
They looked at mutational density and type.

478
00:29:00.588 --> 00:29:03.490
They found a high similarity between the synthetic and the real data.

479
00:29:04.450 --> 00:29:06.091
They looked at mutational signatures.

480
00:29:06.611 --> 00:29:07.532
They found that it's like...

481
00:29:08.568 --> 00:29:09.829
pretty nicely correlated.

482
00:29:10.629 --> 00:29:12.970
The real thing that got me was the genomic distribution.

483
00:29:13.010 --> 00:29:22.054
So you can look at the entire genome of the human from, you know, chromosome one all the way to chromosome 23 with the sex chromosomes.

484
00:29:22.434 --> 00:29:25.615
And you can follow where mutations are happening.

485
00:29:26.055 --> 00:29:26.316
Okay.

486
00:29:26.716 --> 00:29:29.157
And in cancer, mutations do not happen when,

487
00:29:30.317 --> 00:29:30.958
uniformly.

488
00:29:32.239 --> 00:29:38.543
There's going to be parts of the genome where it happens all the time, and there's parts of the genomes where there's very few mutations.

489
00:29:38.843 --> 00:29:43.987
For example, if you're closer to the center of the chromosome, you're not going to get that many mutations.

490
00:29:44.007 --> 00:29:52.193
But if you're closer to the edge, you're going to get a lot more mutations, just because of the physics of like, you know, you got a little bit more freedom towards the edge than towards the center.

491
00:29:52.593 --> 00:29:52.794
Right?

492
00:29:53.614 --> 00:29:57.457
And so they actually showed that mutations in their...

493
00:29:58.527 --> 00:30:02.228
data set would follow non-uniform distribution.

494
00:30:02.248 --> 00:30:08.409
So on the top, they have the distribution of mutations in a real sample.

495
00:30:08.569 --> 00:30:08.769
Yes.

496
00:30:09.129 --> 00:30:11.650
On the bottom, you've got your generated.

497
00:30:11.790 --> 00:30:13.130
It looks pretty much the same.

498
00:30:13.711 --> 00:30:14.431
The eye test.

499
00:30:14.651 --> 00:30:15.491
The eye test.

500
00:30:15.611 --> 00:30:16.771
Yeah, exactly.

501
00:30:17.051 --> 00:30:21.872
And what they've done is like the axis on the x-axis is all the genes.

502
00:30:21.912 --> 00:30:22.253
Yeah, yeah.

503
00:30:22.393 --> 00:30:24.933
Like all 23 chromosomes worth of genomes.

504
00:30:24.973 --> 00:30:25.173
Yeah.

505
00:30:29.435 --> 00:30:32.097
and they've counted how many mutations are happening, right?

506
00:30:32.637 --> 00:30:33.558
That's incredible.

507
00:30:34.178 --> 00:30:38.362
So it's very similar to what we would actually see.

508
00:30:38.862 --> 00:30:47.589
What we've been basically able to do is now create a system that on the fly, it's not fixed.

509
00:30:47.729 --> 00:30:48.990
Yeah.

510
00:30:49.130 --> 00:30:50.551
It's a generative model.

511
00:30:50.731 --> 00:30:50.991
Yeah.

512
00:30:51.151 --> 00:30:52.613
So on the fly...

513
00:30:53.393 --> 00:30:59.758
Any team can now take this and say, we want this kind of dynamics.

514
00:31:00.518 --> 00:31:10.005
And they can generate now these cancer data sets that mimic real cancer genomes to a degree of –

515
00:31:11.541 --> 00:31:20.532
accuracy or comparable high fidelity that is usable in practical work in this space.

516
00:31:20.592 --> 00:31:21.052
Exactly.

517
00:31:21.352 --> 00:31:23.996
And that's that key that you just touched on, usable, right?

518
00:31:24.036 --> 00:31:25.257
What does it mean to be usable?

519
00:31:25.577 --> 00:31:30.543
Well, they actually tried to test it with something called deep tumor.

520
00:31:30.563 --> 00:31:31.404
Deep tumor is a

521
00:31:32.727 --> 00:31:36.568
Again, another AI that's been developed to identify cancers.

522
00:31:36.768 --> 00:31:37.028
Okay?

523
00:31:37.408 --> 00:31:39.689
They could fool deep tumor really well.

524
00:31:39.709 --> 00:31:40.049
Oh, wow.

525
00:31:40.069 --> 00:31:40.829
But here's the key.

526
00:31:40.969 --> 00:31:41.469
Here's the key.

527
00:31:41.969 --> 00:31:46.550
Deep tumor is really bad at rare cancer subtypes.

528
00:31:47.010 --> 00:31:47.210
Okay.

529
00:31:47.230 --> 00:31:47.490
Okay?

530
00:31:47.790 --> 00:31:51.851
Like lymph MCLL, there's only 35 samples in the training set.

531
00:31:52.271 --> 00:31:52.551
Okay?

532
00:31:53.271 --> 00:31:58.553
And if you have low mutations, you're going to struggle with creating an AI that can identify that.

533
00:31:58.833 --> 00:32:00.433
Well, they substituted...

534
00:32:02.058 --> 00:32:06.057
generated samples into the training set, and DeepTumor did better.

535
00:32:07.075 --> 00:32:10.876
Once deep tumor trained on those, the F1 scores improved.

536
00:32:11.016 --> 00:32:12.536
That's fascinating.

537
00:32:12.896 --> 00:32:20.158
The synthetic data from this GAN, this OncoGAN, when fed into an existing detection model.

538
00:32:20.178 --> 00:32:21.338
That's kind of established.

539
00:32:21.458 --> 00:32:22.639
People use it all the time.

540
00:32:22.919 --> 00:32:23.939
It's credible.

541
00:32:23.999 --> 00:32:24.899
All these things.

542
00:32:25.159 --> 00:32:27.800
It has now improved its ability for these edge cases.

543
00:32:27.820 --> 00:32:29.160
These edge niche cases.

544
00:32:29.240 --> 00:32:36.502
Because now I can generate genomes for that specific thing that's not really prevalent.

545
00:32:37.422 --> 00:32:39.083
in a lot of the real world data sets.

546
00:32:39.284 --> 00:32:47.249
Because the idea is you can take the deep tumor and then use it on like real world patient it and know that they do or do not have it.

547
00:32:47.449 --> 00:32:47.749
Exactly.

548
00:32:47.769 --> 00:32:52.332
And then if it comes out with the result, that is really powerful.

549
00:32:52.432 --> 00:33:01.818
It's really augmenting the data sets that we already have with these synthetic data sets, which by the way, this is something that people do all the time in AI research, right?

550
00:33:01.918 --> 00:33:05.120
You do data augmentation where you take whatever data you have

551
00:33:05.460 --> 00:33:07.241
You like apply transforms to it.

552
00:33:07.281 --> 00:33:09.382
When it comes to image data, you like squish it.

553
00:33:09.402 --> 00:33:13.923
You'll like turn it from black and white to red, red, white, and blue or vice versa.

554
00:33:14.224 --> 00:33:19.486
You'll try to make the model more robust to the data.

555
00:33:19.506 --> 00:33:19.626
Yeah.

556
00:33:20.734 --> 00:33:47.301
quality but here they're actually able to create new data that can target these really rare cancer subtypes and establish a new paradigm for medical ai right we can generate more data instead of we just need to collect more data right right right which which you know the there is a bottleneck on the collecting data piece exactly and so you can much more quickly scale uh if you can uh

557
00:33:49.147 --> 00:33:59.097
create usable synthetic data sets that actually map to the real world use cases in a way that is demonstrably true.

558
00:33:59.297 --> 00:33:59.537
Yes.

559
00:33:59.777 --> 00:34:02.080
Which is kind of the point, part of the point of this.

560
00:34:02.220 --> 00:34:03.521
Yeah, exactly, exactly.

561
00:34:03.561 --> 00:34:06.204
And there's no one-to-one correspondence with the real patients.

562
00:34:06.404 --> 00:34:06.644
Right.

563
00:34:06.744 --> 00:34:08.186
You don't have to worry about HIPAA.

564
00:34:08.666 --> 00:34:12.450
You solve the privacy issue while accelerating the ability to...

565
00:34:13.811 --> 00:34:20.433
both create the synthetic data sets and then have those applied to existing tool sets or create new tool sets around all this process.

566
00:34:20.633 --> 00:34:27.795
And the team already, they've got 800 synthetic genomes that are openly available, no ethics approval needed, right?

567
00:34:27.815 --> 00:34:31.716
So anyone can go and try to train a model based on that.

568
00:34:31.736 --> 00:34:37.938
If they've got a new architecture in their mind that really is specific to genomic data, they can test it on this.

569
00:34:38.598 --> 00:34:40.618
It's a new gold standard for benchmarking, right?

570
00:34:40.638 --> 00:34:42.599
Because now when I generate...

571
00:34:43.359 --> 00:34:47.667
some kind of data set, I have a ground truth on what type of cancer it is.

572
00:34:47.767 --> 00:34:49.650
What are the mutations that should be flagged?

573
00:34:49.910 --> 00:34:50.932
So on and so forth.

574
00:34:51.213 --> 00:34:53.877
So it's a catalyst for...

575
00:34:54.912 --> 00:34:56.313
medical oncology, I think.

576
00:34:56.573 --> 00:34:57.514
I think it's very cool.

577
00:34:57.534 --> 00:35:01.657
It's a huge new tool in the toolbox for oncology studies.

578
00:35:01.837 --> 00:35:02.017
Yeah.

579
00:35:02.237 --> 00:35:03.958
I mean, there's still limitations, right?

580
00:35:04.379 --> 00:35:15.286
Like the fact that we're doing this factorization where we first create the mutations and then we have another model that goes in and distributes it across the genome.

581
00:35:15.786 --> 00:35:16.247
It's not...

582
00:35:16.587 --> 00:35:32.727
fully capturing this complex interplay between features like for example if you have a copy number alter alteration where you have these giant chunks of chromosome that get duplicated or deleted that's going to influence local mutation rate this is not something that captures that right?

583
00:35:32.727 --> 00:35:44.471
It can't simulate complex events like chromothripsis and tumor subclones, but it's still a huge, huge step forward.

584
00:35:45.891 --> 00:35:47.051
It's one step at a time.

585
00:35:47.491 --> 00:35:53.573
This has obviously been an issue that is not only affecting so many people, but it's been happening for so long.

586
00:35:53.653 --> 00:35:58.575
And any new step towards...

587
00:35:59.475 --> 00:36:02.797
our ability to solve this more robustly is like hugely important.

588
00:36:02.837 --> 00:36:03.017
Yeah.

589
00:36:03.037 --> 00:36:03.177
Yeah.

590
00:36:03.197 --> 00:36:03.398
Yeah.

591
00:36:03.438 --> 00:36:08.761
Because like cancer is just like, it's just such an insanely difficult problem.

592
00:36:08.841 --> 00:36:09.041
Right.

593
00:36:09.081 --> 00:36:09.281
Right.

594
00:36:09.361 --> 00:36:10.682
It's not a single cause.

595
00:36:10.822 --> 00:36:11.923
Right.

596
00:36:12.143 --> 00:36:13.964
There's no single like panacea.

597
00:36:14.465 --> 00:36:14.685
Right.

598
00:36:14.745 --> 00:36:15.305
Solution.

599
00:36:15.365 --> 00:36:15.685
Yeah.

600
00:36:15.745 --> 00:36:17.246
It's just, yeah.

601
00:36:17.306 --> 00:36:22.570
So, so any, any incremental, and this, this I think is actually a huge step because I think this opens up

602
00:36:23.708 --> 00:36:34.318
other people to do this kind of stuff, to create large data sets, to then have really AI going in full-fledged to try and help out.

603
00:36:34.458 --> 00:36:35.419
Right, right, right.

604
00:36:35.539 --> 00:36:43.646
In a way that, again, keeps in mind some of the aspects of data privacy that are very important when you talk about medical studies.

605
00:36:44.647 --> 00:36:45.808
For all the reasons we talked about,

606
00:36:46.769 --> 00:37:11.835
at the top of the story uh this is actually this is really really again we said we're gonna make sure we give toronto yeah university of toronto i mean their ai is uh this is a big deal their their ai is always like on point you know their ai research is is ridiculous so this is a really big deal this is a really big deal i mean you yeah you guys lost the world series but we'll see you next year hopefully you know that'll be great

607
00:37:12.615 --> 00:37:13.556
You know, round three.

608
00:37:14.317 --> 00:37:14.837
Round three.

609
00:37:14.977 --> 00:37:17.599
You've given us an incredible new tool in the box.

610
00:37:17.619 --> 00:37:18.320
Well done on that.

611
00:37:18.340 --> 00:37:25.806
The ability to do cancer detection and expand the amount of people that can start really working at this issue in earnest.

612
00:37:26.466 --> 00:37:27.547
That's our story number one.

613
00:37:27.647 --> 00:37:27.868
Yep.

614
00:37:28.588 --> 00:37:30.630
Number two, the super story.

615
00:37:31.330 --> 00:37:37.876
The Avengers came together for this wireless retinal chip.

616
00:37:37.896 --> 00:37:37.976
Yep.

617
00:37:39.161 --> 00:37:41.242
That's helping solve for blindness.

618
00:37:41.662 --> 00:37:41.862
Yeah.

619
00:37:41.902 --> 00:37:43.602
This is in the New England Journal of Medicine.

620
00:37:43.622 --> 00:37:52.965
Again, we have Science Corporation, Stanford, UCSF, University of Pittsburgh, and University of Bonn that all collaborated on this story.

621
00:37:52.985 --> 00:37:54.565
You were really excited to talk to me about this story.

622
00:37:54.585 --> 00:38:06.389
I was really excited to talk to you about this because my undergrad PhD thesis, not PhD, my undergrad bachelor's thesis in Princeton had to do with

623
00:38:07.221 --> 00:38:07.741
the retina.

624
00:38:08.442 --> 00:38:18.648
And, you know, we were doing fundamental science, but this is a really cool story because it's, it's out in NBC news, Fox news, even covered it with a tiny eye implant.

625
00:38:19.990 --> 00:38:28.039
and a set of glasses, we can now cure age-related macular degeneration.

626
00:38:28.239 --> 00:38:28.980
Which is crazy.

627
00:38:29.020 --> 00:38:30.001
Which is crazy, right?

628
00:38:30.321 --> 00:38:33.605
The solution is pretty ridiculous.

629
00:38:33.685 --> 00:38:42.434
Now, to be mindful, this age-related macular degeneration, which is basically age-related blindness in the center of your brain,

630
00:38:43.697 --> 00:38:44.658
of your visual field.

631
00:38:45.478 --> 00:38:49.641
This had no prior restorative therapies, right?

632
00:38:49.741 --> 00:38:53.104
This is the first of its kind that is trying to tackle that.

633
00:38:53.204 --> 00:38:53.444
Okay.

634
00:38:53.624 --> 00:39:00.189
So basically if you had AMD, you were just, yeah, you were just, you were just blind in the middle of your, in the middle of your field.

635
00:39:00.209 --> 00:39:03.892
You could see in the peripheral, but in the middle, you couldn't see anything.

636
00:39:04.312 --> 00:39:07.755
And now we actually have something that is restorative.

637
00:39:08.655 --> 00:39:10.137
Not just slowing it down.

638
00:39:10.677 --> 00:39:11.718
Restorative.

639
00:39:11.798 --> 00:39:14.301
It brings it back.

640
00:39:14.321 --> 00:39:16.402
The solution is something called the Prima system.

641
00:39:16.442 --> 00:39:19.365
It's a photovoltaic retina implant microarray.

642
00:39:19.806 --> 00:39:21.747
It's a novel neuroprosthetic.

643
00:39:21.807 --> 00:39:22.728
So it's wireless.

644
00:39:22.988 --> 00:39:23.889
It's subretinal.

645
00:39:24.510 --> 00:39:26.872
And we're going to get into what all that means.

646
00:39:26.892 --> 00:39:29.795
This was published in the New England Journal of Medicine.

647
00:39:30.375 --> 00:39:32.898
The key trial is called Primavera.

648
00:39:33.498 --> 00:39:42.446
I don't know if the authors are fans of the music festival in Barcelona, but that's what they called it, the Primavera.

649
00:39:42.586 --> 00:39:47.991
It's an incredible thing, okay?

650
00:39:48.011 --> 00:39:54.497
81% of participants said they got profound improvement, meaningful improvement, okay?

651
00:39:54.837 --> 00:39:56.098
And remember those like...

652
00:39:58.316 --> 00:40:03.042
eye exams that you had to do with like a bunch of big, big like letters.

653
00:40:03.222 --> 00:40:09.149
And then these guys, the guys who had this over five additional lines on the eye chart.

654
00:40:09.530 --> 00:40:10.150
Really?

655
00:40:10.230 --> 00:40:10.491
Right.

656
00:40:11.312 --> 00:40:13.134
That's, that's, that's a lot.

657
00:40:13.294 --> 00:40:15.697
I have bad astigmatism and can barely see.

658
00:40:15.817 --> 00:40:16.778
And so I, I,

659
00:40:18.039 --> 00:40:22.761
And some patients regain the ability to read letters, numbers, and words.

660
00:40:23.861 --> 00:40:24.722
That's crazy.

661
00:40:24.762 --> 00:40:24.962
Okay?

662
00:40:24.982 --> 00:40:30.964
So let's talk about how the eye works because that's going to tell you about how the mechanism of this thing works.

663
00:40:31.024 --> 00:40:31.244
Okay?

664
00:40:31.584 --> 00:40:33.324
The eye is basically a camera.

665
00:40:33.945 --> 00:40:34.205
Okay?

666
00:40:35.165 --> 00:40:36.065
You've got a lens.

667
00:40:37.026 --> 00:40:37.506
You've got...

668
00:40:38.700 --> 00:40:40.861
a detector in the back, just like a camera.

669
00:40:40.881 --> 00:40:51.388
A camera has a lens and then a CCD in the back that the image falls on, and then the CCD captures how many photons are coming here versus there, what are the colors of those things.

670
00:40:51.808 --> 00:40:53.689
The eye is very, very similar, okay?

671
00:40:53.709 --> 00:40:56.891
The retina is an image sensor in the back.

672
00:40:57.372 --> 00:41:02.155
And this retina has photoreceptors, which are your rods and cones.

673
00:41:02.195 --> 00:41:03.696
You can think of that as your CCD.

674
00:41:04.036 --> 00:41:05.977
That's the thing that actually interacts with the light.

675
00:41:06.357 --> 00:41:06.958
And then if...

676
00:41:07.798 --> 00:41:11.720
a little photon interacts with it, it's going to set off a little electrical signal.

677
00:41:12.100 --> 00:41:15.461
Then it goes through a bunch of different cells.

678
00:41:16.302 --> 00:41:17.782
They're called bipolar cells.

679
00:41:17.843 --> 00:41:19.543
Some of them are called amicron cells.

680
00:41:20.904 --> 00:41:27.867
And at the end of the day, you've got ganglion cells, which are the cells at the very end that...

681
00:41:28.762 --> 00:41:40.130
take in all of that light information, convert it into electrical pulses, action potentials, and that goes through the optic nerve, through your eye, to your brain, to the occipital lobe.

682
00:41:40.570 --> 00:41:45.394
And that's what's transmitting visual information to the brain's visual cortex, okay?

683
00:41:45.674 --> 00:41:50.217
But it's really, it's like mechanistically, the eye is remarkably like a camera.

684
00:41:50.477 --> 00:41:53.760
You've got a lens, and that lens you can squish

685
00:41:54.520 --> 00:42:00.604
There's muscles on the end of your lens that you can squish and so on and so forth to let in more light, less light.

686
00:42:00.644 --> 00:42:03.286
That's your aperture, right?

687
00:42:03.586 --> 00:42:07.529
In a normal DSLR, you've got the magnifying, right?

688
00:42:07.789 --> 00:42:17.255
And the magnification is like how much... Like the distance between your lens and the focal plane.

689
00:42:17.275 --> 00:42:17.875
Yes, yes.

690
00:42:18.155 --> 00:42:21.558
But for us, I can focus on the...

691
00:42:22.316 --> 00:42:23.518
microphone right in front of me.

692
00:42:24.018 --> 00:42:27.483
And I can also focus on a mountain that's way, way, way far away.

693
00:42:27.984 --> 00:42:28.204
Right.

694
00:42:28.324 --> 00:42:28.625
Yeah.

695
00:42:28.725 --> 00:42:30.748
And the way that our, I can do that.

696
00:42:30.828 --> 00:42:33.611
We don't have the luxury of actually extending our lens.

697
00:42:34.072 --> 00:42:37.417
So what we do is we actually change the curvature of the lens lens.

698
00:42:38.246 --> 00:42:38.826
In order to do so.

699
00:42:38.886 --> 00:42:41.848
In order to actually get different focal lengths, right?

700
00:42:41.868 --> 00:42:42.088
I get you.

701
00:42:42.148 --> 00:42:45.029
So it's remarkably like a camera, okay?

702
00:42:45.469 --> 00:42:55.973
And in the back of your camera, in the back of the eye, you have the retina, which is this mini computer that takes in all of that information and then relays that back to the brain.

703
00:42:56.053 --> 00:42:57.614
It's sort of like the processing center.

704
00:42:57.834 --> 00:42:58.474
Yes.

705
00:42:58.594 --> 00:43:02.376
Once the raw data comes in, it kind of packages it up to send it upstairs.

706
00:43:02.416 --> 00:43:03.817
To send it upstairs, exactly.

707
00:43:04.277 --> 00:43:04.437
And...

708
00:43:05.428 --> 00:43:09.509
In the center of your retina, you have something called the macula and the fovea.

709
00:43:09.729 --> 00:43:10.009
Okay?

710
00:43:10.589 --> 00:43:14.350
Notice, like, I want you to just do this experiment where you're, like, looking at stuff.

711
00:43:15.190 --> 00:43:20.532
Notice that you have a lot of detail in the center of your vision compared to the periphery.

712
00:43:20.652 --> 00:43:20.972
Yeah.

713
00:43:21.152 --> 00:43:21.392
Right?

714
00:43:21.432 --> 00:43:21.692
Yeah.

715
00:43:21.912 --> 00:43:24.833
Like, if you really wanted to read something, like...

716
00:43:25.553 --> 00:43:31.235
that stuff that's over there, you would focus on it so that it's at the center of your vision.

717
00:43:31.695 --> 00:43:34.056
If you focus somewhere else, you wouldn't be able to read it.

718
00:43:34.636 --> 00:43:47.941
And the reason for that is you've got this structure called the macula endophovia that has a high density of rods and cones, specifically cones, actually, that are the color receptors.

719
00:43:48.921 --> 00:43:50.002
So you've got like...

720
00:43:51.795 --> 00:43:55.479
Higher resolution there because the pixel density is higher.

721
00:43:55.639 --> 00:43:59.143
We're not... Everything is not totally in focus all of the time.

722
00:43:59.163 --> 00:44:00.004
Yes, exactly.

723
00:44:00.424 --> 00:44:06.911
And macular degeneration is actually a progressive disease that damages that macula.

724
00:44:06.931 --> 00:44:07.572
So...

725
00:44:07.852 --> 00:44:10.154
The macula is in the middle, right?

726
00:44:10.234 --> 00:44:15.319
And you can have two types of AMD, age-related macular degeneration.

727
00:44:15.359 --> 00:44:25.309
There's the wet AMD where basically the abnormal blood vessel growth, there's a bunch of blood vessels in your eyes, right, that are keeping these cells alive, giving it oxygen, things like that.

728
00:44:25.629 --> 00:44:28.051
Those blood vessels can leak, rapid vision loss.

729
00:44:29.001 --> 00:44:32.803
And then there's dry atrophic AMD, okay?

730
00:44:33.083 --> 00:44:35.464
And that's the gradual thinning of the macula.

731
00:44:35.884 --> 00:44:42.668
You get these extracellular deposits where the cells are sort of just dumping their trash, and then that kills everything.

732
00:44:43.168 --> 00:44:49.851
And all of that accumulation happens, and it kills all of those cones that are there.

733
00:44:49.931 --> 00:44:50.392
In that center.

734
00:44:50.412 --> 00:44:51.012
In that center.

735
00:44:51.272 --> 00:44:56.915
And so that center of your, the center of your visual field Yes.

736
00:44:57.650 --> 00:45:01.456
is the thing that goes away, which is like, obviously, and that's the whole point.

737
00:45:01.476 --> 00:45:01.856
Yeah.

738
00:45:01.976 --> 00:45:02.217
Right.

739
00:45:02.257 --> 00:45:02.517
Right.

740
00:45:02.557 --> 00:45:02.818
Right.

741
00:45:02.858 --> 00:45:06.703
That's, that's where most of my information is coming from.

742
00:45:06.863 --> 00:45:07.044
Right.

743
00:45:07.504 --> 00:45:09.107
And that's the one that's more, more common.

744
00:45:09.267 --> 00:45:09.467
Right.

745
00:45:09.487 --> 00:45:09.828
Okay.

746
00:45:09.888 --> 00:45:11.710
And then there's geographic atrophy.

747
00:45:12.512 --> 00:45:12.692
Um,

748
00:45:12.912 --> 00:45:15.773
which it's at the end stage of AMD.

749
00:45:16.153 --> 00:45:20.395
You get these well demarcated patches of your retinal tissue where they die off.

750
00:45:20.875 --> 00:45:22.956
You get the photoreceptors that are dying off.

751
00:45:23.697 --> 00:45:28.018
And it results in like absolute scotoma.

752
00:45:28.198 --> 00:45:30.859
So you get a blind spot right in the middle.

753
00:45:31.080 --> 00:45:36.682
And when you get the blind spot, you literally can't see in the middle of your visual field, which is so crazy, right?

754
00:45:37.102 --> 00:45:40.083
It's a complete loss of phototransduction.

755
00:45:40.143 --> 00:45:40.784
Okay.

756
00:45:40.804 --> 00:45:41.024
Okay.

757
00:45:42.004 --> 00:45:46.008
That is what this particular study is trying to handle.

758
00:45:46.308 --> 00:45:46.569
Okay?

759
00:45:47.429 --> 00:45:49.552
And it's pretty common, actually.

760
00:45:49.892 --> 00:46:00.662
Irreversible blindness among older patients globally, you've got something like 200 million people that have this.

761
00:46:01.123 --> 00:46:03.025
In the U.S. itself, it's a million people.

762
00:46:04.853 --> 00:46:08.916
It's an incredibly big problem that we need to solve, right?

763
00:46:08.996 --> 00:46:11.298
Obviously, quality of life goes away.

764
00:46:11.678 --> 00:46:13.020
There's a loss of independence.

765
00:46:13.380 --> 00:46:17.783
It's got a devastating impact on all aspects of your life.

766
00:46:17.803 --> 00:46:20.225
Especially when... Not that this is...

767
00:46:20.625 --> 00:46:21.186
It's true.

768
00:46:21.426 --> 00:46:22.887
It's been true forever.

769
00:46:23.307 --> 00:46:26.230
But particularly when everything is so device-driven.

770
00:46:26.770 --> 00:46:27.791
And, you know...

771
00:46:28.291 --> 00:46:28.811
Oh, yeah.

772
00:46:28.991 --> 00:46:29.752
I didn't even think of that.

773
00:46:29.772 --> 00:46:30.232
You know what I mean?

774
00:46:30.252 --> 00:46:34.913
So it's like you have to be able to read and interact with screens.

775
00:46:35.894 --> 00:46:37.334
And so it's impossible to do so.

776
00:46:37.674 --> 00:46:38.435
Exactly.

777
00:46:38.995 --> 00:46:41.636
Especially in a device-driven kind of era.

778
00:46:41.836 --> 00:46:42.316
Yeah, yeah.

779
00:46:42.456 --> 00:46:47.178
And, you know, previously it was just like, oh, you're just – you're totally –

780
00:46:47.958 --> 00:46:48.979
not having a good time.

781
00:46:49.219 --> 00:46:49.459
Okay.

782
00:46:49.759 --> 00:46:50.940
There are a few drugs.

783
00:46:51.360 --> 00:46:54.562
Like there's a, there's the first drug was approved in 2023.

784
00:46:54.582 --> 00:46:55.703
I survey.

785
00:46:55.723 --> 00:47:06.390
Um, there's another drug called, uh, syphovray, I guess the mechanism is you basically inhibit this cascade of, uh,

786
00:47:07.050 --> 00:47:09.551
Stuff that's going wrong in the back of your retina.

787
00:47:09.991 --> 00:47:12.432
It slows the lesion growth by 20%.

788
00:47:12.732 --> 00:47:15.272
But at the end of the day, it's not restorative.

789
00:47:15.312 --> 00:47:16.393
It's just stopping.

790
00:47:16.413 --> 00:47:17.753
It's just slowing it down.

791
00:47:17.793 --> 00:47:18.533
The degeneration.

792
00:47:18.573 --> 00:47:18.813
Right?

793
00:47:18.914 --> 00:47:20.014
And that's great.

794
00:47:20.554 --> 00:47:20.774
Yes.

795
00:47:21.154 --> 00:47:23.615
But for someone who's already had it,

796
00:47:24.679 --> 00:47:25.640
It's not really doing much.

797
00:47:25.660 --> 00:47:27.021
Yeah, right, right, right.

798
00:47:27.161 --> 00:47:27.341
Right?

799
00:47:27.401 --> 00:47:29.163
And so now we've got this Prima system.

800
00:47:29.203 --> 00:47:31.525
It's a new paradigm in vision restoration.

801
00:47:31.865 --> 00:47:33.326
It is restoring things, okay?

802
00:47:33.346 --> 00:47:35.268
It's got two primary parts.

803
00:47:35.628 --> 00:47:37.029
There's an internal component.

804
00:47:37.090 --> 00:47:41.073
There's a tiny, tiny photovoltaic implant, okay?

805
00:47:41.413 --> 00:47:44.896
It's a little cluster of neuro...

806
00:47:46.137 --> 00:47:47.519
implants, right?

807
00:47:47.659 --> 00:47:52.586
It's a tiny little electronic device at the end of the day that goes in the back of your head, okay?

808
00:47:53.227 --> 00:47:57.313
And it surgically gets placed in the back of your retina, in the back of your eye.

809
00:47:57.653 --> 00:48:01.478
And then there's an external component, which is specialized glasses that

810
00:48:02.007 --> 00:48:03.188
with a video camera.

811
00:48:03.689 --> 00:48:22.211
And what that video camera is going to do is project light through your eye to that implant in the back of your retina and stimulate that implant to create electrical signals to trick your eye into thinking your photoreceptors are still alive.

812
00:48:23.398 --> 00:48:24.839
That's – okay.

813
00:48:24.859 --> 00:48:34.988
So the idea is you have this two-part system which is effectively mimicking the real effect of photons hitting the back of your retina.

814
00:48:35.308 --> 00:48:42.134
But because all the stuff in the macula is now dead, we're not like regenerating the cells itself.

815
00:48:42.654 --> 00:48:51.042
But we are creating a synthetic version of the exact same input such that when it goes into your brain, it's receiving –

816
00:48:51.622 --> 00:48:56.324
an identical signal to what you would in a natural sense, but in an artificial...

817
00:48:57.184 --> 00:48:58.144
Very interesting.

818
00:48:58.324 --> 00:48:58.704
Exactly.

819
00:48:58.784 --> 00:49:02.386
It's like simple, but at the same time, extremely hard to do.

820
00:49:02.506 --> 00:49:03.686
Obviously, because it hasn't been done yet.

821
00:49:03.766 --> 00:49:06.647
Yeah, and it's only being done because of today's technology.

822
00:49:06.667 --> 00:49:09.948
So now, let's get into... First, let's tackle the implant.

823
00:49:09.988 --> 00:49:12.769
What is the thing that is going in the back of your eye?

824
00:49:13.309 --> 00:49:17.130
It's a 2 nanometer by 2 nanometer little chip.

825
00:49:17.570 --> 00:49:18.671
30 microns thick.

826
00:49:19.331 --> 00:49:20.291
It's a silicon chip.

827
00:49:20.591 --> 00:49:28.014
It's a pixel array of 378 different independent hexagonal chip pixels, two nanometers by two nanometers.

828
00:49:28.134 --> 00:49:30.315
It's way, way smaller than a penny.

829
00:49:30.635 --> 00:49:34.096
It's like, it's like, you know, smaller than even Lincoln's.

830
00:49:34.176 --> 00:49:39.858
It's like the size of Lincoln's eye on the pet on the penny in the, in the, you know, memorial in the, yeah.

831
00:49:39.878 --> 00:49:42.059
That's how small this thing is, right?

832
00:49:42.079 --> 00:49:42.299
Small.

833
00:49:45.457 --> 00:49:53.144
Each of the pixels, there's 378 pixels, each of those pixels has a little mini solar panel and an electrode.

834
00:49:53.464 --> 00:49:55.126
Now you're thinking, why solar panels?

835
00:49:55.546 --> 00:49:59.189
Well, solar panels create electricity out of light.

836
00:50:00.510 --> 00:50:05.335
Now, I've created electricity from the light that is going in.

837
00:50:06.710 --> 00:50:07.551
Into my eye.

838
00:50:07.871 --> 00:50:10.634
I don't need an external battery or anything like that.

839
00:50:10.874 --> 00:50:12.015
I'm so mad.

840
00:50:12.175 --> 00:50:14.337
I thought that was so cool, dude.

841
00:50:14.357 --> 00:50:15.298
That's so genius.

842
00:50:15.958 --> 00:50:17.580
I thought that was so cool.

843
00:50:17.980 --> 00:50:22.004
Which matters because a lot of times you can even use phones as an example.

844
00:50:22.384 --> 00:50:25.767
A lot of the weight that goes into the phone is the battery.

845
00:50:26.308 --> 00:50:28.049
This is the first one that is wireless.

846
00:50:28.189 --> 00:50:32.053
The reason why it's wireless is because they use this photovoltaic cell.

847
00:50:32.153 --> 00:50:32.674
That makes sense.

848
00:50:33.586 --> 00:50:34.247
That's genius.

849
00:50:34.307 --> 00:50:35.768
And I think that's so genius.

850
00:50:35.788 --> 00:50:36.348
Yeah, yeah.

851
00:50:36.408 --> 00:50:36.708
Right?

852
00:50:36.989 --> 00:50:48.997
Like there's an electrode and that electrode is what's actually creating the electric field to then tag all the other neurons in the retina to make it seem like, right, there is a photoreceptor there.

853
00:50:49.077 --> 00:50:50.178
There's a rod and cone there.

854
00:50:50.678 --> 00:50:55.304
But the energy for that electricity is coming from the photon itself.

855
00:50:55.585 --> 00:50:57.427
And it makes sense because your eyes are, anyway.

856
00:50:57.888 --> 00:50:58.148
Right?

857
00:50:58.328 --> 00:50:59.249
Yeah, that's smart.

858
00:50:59.370 --> 00:51:00.231
That's so smart.

859
00:51:00.251 --> 00:51:01.052
That's very smart.

860
00:51:01.072 --> 00:51:02.033
That's so smart.

861
00:51:02.153 --> 00:51:03.375
I love that.

862
00:51:03.515 --> 00:51:06.098
And as soon as I saw it, I was like, oh, that's dope.

863
00:51:08.520 --> 00:51:08.980
That's dope.

864
00:51:09.080 --> 00:51:09.400
That's good.

865
00:51:09.420 --> 00:51:09.580
Yeah.

866
00:51:09.780 --> 00:51:15.462
And so what they do is they take this little chip, two nanometers by two nanometers, and they put it under the retina.

867
00:51:15.622 --> 00:51:15.862
Okay.

868
00:51:15.882 --> 00:51:19.883
This is deliberate where the electrodes are in proximity to the bipolar cells.

869
00:51:19.903 --> 00:51:23.805
So it's, it's where the, it's where the photoreceptors would be.

870
00:51:23.885 --> 00:51:30.827
And there you can see on the left, you've got a giant patch of where the photoreceptors have died.

871
00:51:31.207 --> 00:51:34.228
You're putting that chip right underneath.

872
00:51:34.548 --> 00:51:34.688
Mm-hmm.

873
00:51:35.030 --> 00:51:36.732
Where those photoreceptors should be.

874
00:51:37.493 --> 00:51:37.793
Okay?

875
00:51:38.213 --> 00:51:48.545
And that chip is now going to stimulate those bipolar cells, which will then in turn stimulate the ganglion cells, which will then go all the way to the back of your brain.

876
00:51:48.825 --> 00:51:49.005
Right?

877
00:51:49.285 --> 00:51:50.146
It's natural.

878
00:51:50.647 --> 00:51:52.249
It's efficiently processed signal.

879
00:51:52.529 --> 00:51:55.893
And it's more sophisticated than these older retinal implants.

880
00:51:56.934 --> 00:52:01.135
That we're stimulating the ganglion cells that were directly going into the brain.

881
00:52:01.495 --> 00:52:03.816
These guys are mimicking the photoreceptors.

882
00:52:03.916 --> 00:52:04.436
Itself.

883
00:52:04.496 --> 00:52:05.156
Yeah.

884
00:52:05.356 --> 00:52:05.716
You see?

885
00:52:05.776 --> 00:52:06.357
Yeah.

886
00:52:06.477 --> 00:52:08.937
It's so nice, dude.

887
00:52:09.477 --> 00:52:10.098
I love that.

888
00:52:10.378 --> 00:52:16.299
And what I love about this as well is that the pixel has an input.

889
00:52:17.159 --> 00:52:21.140
So that's where the source of the electric field, like the source of the current.

890
00:52:21.521 --> 00:52:22.761
But it's also got a sync.

891
00:52:24.025 --> 00:52:26.947
Because if you don't have a sink, then you put a lot of current out.

892
00:52:27.227 --> 00:52:33.611
That current is going to spread out and it's going to trigger all the rest of the cells everywhere else in the retina.

893
00:52:33.871 --> 00:52:36.152
But if you have a sink, then it becomes localized.

894
00:52:36.172 --> 00:52:36.973
The current goes out.

895
00:52:37.413 --> 00:52:40.795
It triggers the stuff in its neighborhood and it goes back in.

896
00:52:40.815 --> 00:52:42.036
It becomes discreet.

897
00:52:42.176 --> 00:52:43.557
Yes, exactly.

898
00:52:43.777 --> 00:52:44.157
Exactly.

899
00:52:44.197 --> 00:52:48.160
So now your pixel size, right, for like what I'm trying to see, that becomes smaller.

900
00:52:48.180 --> 00:52:48.320
Yeah.

901
00:52:50.083 --> 00:52:50.704
They've thought of it.

902
00:52:50.784 --> 00:52:51.864
They've thought of everything, mate.

903
00:52:51.884 --> 00:52:52.545
This is really good.

904
00:52:52.565 --> 00:52:53.466
It's really nice.

905
00:52:53.486 --> 00:52:54.126
This is really good.

906
00:52:54.446 --> 00:52:55.687
Okay, so that's the implant.

907
00:52:55.848 --> 00:52:56.128
Okay.

908
00:52:56.168 --> 00:52:57.729
Now let's get into the outside, right?

909
00:52:57.789 --> 00:52:57.989
Yes.

910
00:52:58.610 --> 00:53:02.593
There needs to be something that actually goes and puts light to that implant.

911
00:53:03.744 --> 00:53:05.847
chip that's in the back of your retina.

912
00:53:05.907 --> 00:53:06.148
Yes.

913
00:53:06.208 --> 00:53:06.368
Right?

914
00:53:06.568 --> 00:53:07.970
And that's where these glasses come in.

915
00:53:08.010 --> 00:53:09.192
So there's a user interface.

916
00:53:09.212 --> 00:53:10.294
There's a smart component.

917
00:53:10.654 --> 00:53:11.756
You put on these glasses.

918
00:53:12.056 --> 00:53:13.438
You also have a user interface.

919
00:53:13.478 --> 00:53:14.820
Like it's kind of like a remote.

920
00:53:15.841 --> 00:53:19.807
And that's your user interface that you can like zoom in.

921
00:53:20.528 --> 00:53:21.349
Zoom out.

922
00:53:22.270 --> 00:53:22.591
Yeah.

923
00:53:22.611 --> 00:53:26.335
You can adjust the contrast and things like that.

924
00:53:26.636 --> 00:53:31.521
And then your glasses, they've got a little video camera, kind of like the meta glasses.

925
00:53:31.541 --> 00:53:33.384
You know, the meta glasses have a video camera.

926
00:53:33.784 --> 00:53:38.229
And those glasses then send in near-infrared.

927
00:53:39.479 --> 00:53:43.762
signals to the back of your retina, which is where the chip is.

928
00:53:44.202 --> 00:53:45.242
Why near-infrared?

929
00:53:45.282 --> 00:53:51.326
Well, all the other natural photoreceptors in your eye are not sensitive to near-infrared.

930
00:53:52.347 --> 00:53:52.807
Right?

931
00:53:52.867 --> 00:53:57.870
If you were sending invisible stuff, then you'd be blinded by like a giant...

932
00:53:58.710 --> 00:54:14.316
like flashlight that's like right next to your eye but because it's near infrared the only thing that is getting stimulated is the chip the chip itself that's that's actually that's really good no that's that's that's really good

933
00:54:14.576 --> 00:54:16.037
It's so nice.

934
00:54:16.077 --> 00:54:17.597
That's, I'm, I'm.

935
00:54:17.877 --> 00:54:18.217
Right?

936
00:54:18.337 --> 00:54:19.278
Yeah, yeah.

937
00:54:19.518 --> 00:54:21.258
You're seeing why I was like decided about this.

938
00:54:21.278 --> 00:54:22.239
This is so cool.

939
00:54:22.559 --> 00:54:25.900
I just, I'm like, I don't have anything to say because it's just, it's so good.

940
00:54:26.020 --> 00:54:26.760
It's so good.

941
00:54:26.800 --> 00:54:27.841
Like they thought of it, right?

942
00:54:28.161 --> 00:54:28.841
It's so good.

943
00:54:28.861 --> 00:54:35.523
The fact that it doesn't have any cables, like usually these things used to come with cables, but the fact that it doesn't have any cables simplifies the surgery.

944
00:54:35.824 --> 00:54:37.664
So you can go into the corner of your eye.

945
00:54:37.744 --> 00:54:39.225
I mean, it's still a crazy surgery.

946
00:54:39.305 --> 00:54:39.785
I'll be honest.

947
00:54:40.105 --> 00:54:44.731
Like you're going in with a needle in the corner of your eye, but you're depositing it in the back of your eye.

948
00:54:44.771 --> 00:54:53.923
But it's a relatively simple surgery because all you do is you stick it in, you leave the little bit of chip where you want it to be, and then you take the needle out, right?

949
00:54:56.947 --> 00:55:00.131
And it's actually incredible.

950
00:55:00.672 --> 00:55:06.480
This chip efficiently absorbs 30 microns of silicon photodiodes, right?

951
00:55:06.700 --> 00:55:09.764
It's 30 microns of silicon photodiodes.

952
00:55:10.105 --> 00:55:12.869
You get this pulse light that's coming in at 30 hertz.

953
00:55:14.090 --> 00:55:21.555
And 30 hertz is, you must know, 30 hertz is about where film frame rate is.

954
00:55:21.615 --> 00:55:27.319
And that's because the ganglion cells that are going to the back of your head, right, that are relaying this information.

955
00:55:27.859 --> 00:55:33.843
Anything faster than 30 hertz, and it's really not discernible, anything slower in it would look like stop motion.

956
00:55:34.003 --> 00:55:39.166
So at 30 hertz, you're getting this flicker-free image of pulse light that's going in.

957
00:55:40.267 --> 00:55:42.289
This is really clever.

958
00:55:42.349 --> 00:55:43.210
Bro, it's so cool.

959
00:55:43.250 --> 00:55:51.738
This is really, really – because they sort of – because you have to both solve for the like technical – so there's a variety of stages of this process, right?

960
00:55:52.059 --> 00:55:58.285
There's the actual like where in the eye do you make the intervention in order to get the restorative benefit?

961
00:55:58.665 --> 00:56:04.030
They made that sort of innovation by putting it in that like where the cones and rods are.

962
00:56:04.090 --> 00:56:04.331
Yeah.

963
00:56:04.951 --> 00:56:07.932
Then you have to think about how do you do, how do you get power?

964
00:56:08.112 --> 00:56:08.653
Yeah.

965
00:56:08.833 --> 00:56:09.453
That's the solar piece.

966
00:56:09.493 --> 00:56:11.674
That's the, yeah, you made a solar cell.

967
00:56:11.694 --> 00:56:12.394
So brilliant.

968
00:56:12.794 --> 00:56:16.096
Then you have to think about like, how are you actually going to turn it on?

969
00:56:16.356 --> 00:56:20.157
Meaning like, how are you going to direct the light to make it actually work the way you want it to?

970
00:56:20.537 --> 00:56:21.398
But not affect everything.

971
00:56:21.418 --> 00:56:22.498
But not affect everything else.

972
00:56:22.978 --> 00:56:26.900
And so the near infrared as the methodology to do so is so genius.

973
00:56:27.000 --> 00:56:27.600
Yeah, dude.

974
00:56:28.001 --> 00:56:29.161
Oh man, that's so good.

975
00:56:29.401 --> 00:56:30.161
It's so good.

976
00:56:30.181 --> 00:56:31.122
That's quite, that's quite nice.

977
00:56:31.262 --> 00:56:32.162
It's so nice.

978
00:56:32.182 --> 00:56:32.702
That's quite nice.

979
00:56:32.722 --> 00:56:33.523
Yeah, it's so nice.

980
00:56:34.243 --> 00:56:39.006
And let's talk about the study itself that came out in the New England Journal of Medicine, right?

981
00:56:39.286 --> 00:56:40.167
It's a multi-center.

982
00:56:40.187 --> 00:56:43.028
So 17 sites across five European countries.

983
00:56:43.328 --> 00:56:44.789
You mentioned the University of Bonn.

984
00:56:45.269 --> 00:56:49.572
I mentioned them specifically because they kind of did a lot of the groundwork.

985
00:56:49.712 --> 00:56:54.995
But a bunch of other European institutions were part of the clinical trial, right?

986
00:56:55.015 --> 00:56:55.816
Sure, sure, sure.

987
00:56:55.876 --> 00:56:59.418
But you're increasing generalizability because of these 17 different studies.

988
00:56:59.978 --> 00:57:01.019
places are trying it.

989
00:57:01.059 --> 00:57:05.343
Because people culturally, genetically, you want to have it spread across a variety of years.

990
00:57:05.363 --> 00:57:05.844
Exactly, yeah.

991
00:57:05.924 --> 00:57:12.550
So 38 participants, all of them over the age of 60, and all of them confirmed atrophy in both eyes.

992
00:57:12.590 --> 00:57:12.951
Both eyes.

993
00:57:13.151 --> 00:57:13.391
Right?

994
00:57:13.431 --> 00:57:13.551
Right.

995
00:57:14.520 --> 00:57:21.746
And what they looked at was something called the log mar, which is the logarithm of minimal angle of resolution.

996
00:57:22.047 --> 00:57:28.412
It's effectively saying like in the center of your field of vision, like how, how big can you see things?

997
00:57:28.512 --> 00:57:29.353
It's this thing.

998
00:57:29.413 --> 00:57:35.838
It's the, it's the letters and the driving test, you know, when you go to the DMV and you're like, okay, can you see, or are you blind?

999
00:57:36.379 --> 00:57:39.942
Those letters tell you how small of a letter you can discern.

1000
00:57:40.422 --> 00:57:41.804
What line can you see on this one?

1001
00:57:43.144 --> 00:57:50.466
I mean, the screen is pretty small, but I think I can see that.

1002
00:57:50.506 --> 00:58:02.389
But in any case, these guys, after 12-month follow-up, after getting the implant, 81%, so 26 out of the 32 participants met the endpoint.

1003
00:58:03.049 --> 00:58:04.069
Highly significant.

1004
00:58:04.809 --> 00:58:07.470
And it's equivalent to five lines.

1005
00:58:07.530 --> 00:58:07.990
Right.

1006
00:58:09.210 --> 00:58:23.683
worth of stuff 20 59 more letters than they could have originally in that eye test, you know, I mean, this is, this is such a huge, I mean, we talk about this all the time, which is like why, you know, um, there's, these studies are so important.

1007
00:58:23.703 --> 00:58:35.773
Like the unlock that you get from something like this, um, for a huge population of people who otherwise we're just going to lose the ability to see.

1008
00:58:36.014 --> 00:58:36.174
Yeah.

1009
00:58:36.194 --> 00:58:36.294
Yeah.

1010
00:58:37.223 --> 00:58:42.094
permanently is now no longer like necessarily true.

1011
00:58:42.255 --> 00:58:42.676
Yeah, dude.

1012
00:58:42.896 --> 00:58:47.066
There was one dude who went from legal blindness to like, I can drive now.

1013
00:58:47.756 --> 00:58:48.176
That's crazy.

1014
00:58:48.196 --> 00:58:49.418
I mean, that's one dude, but still.

1015
00:58:51.019 --> 00:58:53.202
And this is like state, this is V1.

1016
00:58:53.322 --> 00:58:53.582
Yeah.

1017
00:58:53.722 --> 00:58:53.902
Right.

1018
00:58:53.922 --> 00:58:54.022
Yeah.

1019
00:58:54.042 --> 00:58:54.843
This is V1.

1020
00:58:54.903 --> 00:59:08.677
It's the first iteration of this in which you now have, I think that the innovation is in like all the kind of steps we just talked about, because now you can iterate on each of those component parts.

1021
00:59:08.697 --> 00:59:08.998
Yeah.

1022
00:59:09.710 --> 00:59:14.273
For any number of different permutations or use cases, etc.

1023
00:59:14.974 --> 00:59:15.494
Exactly.

1024
00:59:15.594 --> 00:59:16.855
That's a really, really interesting...

1025
00:59:16.875 --> 00:59:17.595
It's really cool.

1026
00:59:17.716 --> 00:59:19.297
I mean, there were detractors.

1027
00:59:19.357 --> 00:59:22.579
Like, if you looked at user satisfaction, 69%.

1028
00:59:22.979 --> 00:59:28.163
So 22 out of the 32 reported medium to high overall satisfaction, which...

1029
00:59:29.160 --> 00:59:35.463
It's still like if you weren't able to see and now you're able to see like that's 30% being like I'm not satisfied.

1030
00:59:35.803 --> 00:59:36.923
And I looked into more of that.

1031
00:59:37.243 --> 00:59:42.625
I think part of that is because of these things called serious adverse events that happen at the end of any surgery.

1032
00:59:42.906 --> 00:59:46.007
Like you're, look, you're sticking a needle into someone's eye.

1033
00:59:46.407 --> 00:59:50.669
or putting in a chip, like there's going to be serious adverse events, right?

1034
00:59:51.070 --> 00:59:55.872
And most of the stuff had to do with the trauma and healing from the surgery.

1035
00:59:55.972 --> 00:59:58.133
Not from the actual device, right?

1036
00:59:58.173 --> 01:00:00.875
So if we can make that process better...

1037
01:00:02.377 --> 01:00:04.860
perhaps this can be a lot better.

1038
01:00:04.880 --> 01:00:05.561
The satisfaction can be higher.

1039
01:00:05.581 --> 01:00:07.183
Yeah, the satisfaction can be higher, right?

1040
01:00:07.203 --> 01:00:07.844
Which makes sense.

1041
01:00:08.044 --> 01:00:08.385
Yeah.

1042
01:00:08.405 --> 01:00:09.125
That does make sense.

1043
01:00:09.286 --> 01:00:15.253
But at the end of the day, I think this Prima is a quantum leap in terms of retinal implants.

1044
01:00:15.594 --> 01:00:20.800
Before, the first FDA-approved retinal implant was something called the Argus 2.

1045
01:00:20.820 --> 01:00:21.361
Okay.

1046
01:00:21.661 --> 01:00:23.443
Which had 60 electrodes.

1047
01:00:23.463 --> 01:00:25.384
That was this image I just pulled up earlier.

1048
01:00:25.424 --> 01:00:27.846
Yeah, and that was wired, though.

1049
01:00:27.886 --> 01:00:28.046
Right.

1050
01:00:28.407 --> 01:00:30.468
You can see it coming out of the... Yeah, yeah.

1051
01:00:30.568 --> 01:00:34.171
And you can see that in the retina itself, there's a little wire that are like...

1052
01:00:34.892 --> 01:00:35.092
Right?

1053
01:00:35.152 --> 01:00:36.433
This thing is just complete...

1054
01:00:36.453 --> 01:00:38.295
It's just a chip naked on its own.

1055
01:00:38.335 --> 01:00:39.395
Wireless.

1056
01:00:39.456 --> 01:00:40.476
Wireless, dude.

1057
01:00:41.057 --> 01:00:42.458
Wireless, no battery.

1058
01:00:42.518 --> 01:00:43.219
No battery.

1059
01:00:43.379 --> 01:00:44.840
It's just a photovoltaic cell.

1060
01:00:46.943 --> 01:00:47.944
It's so clever.

1061
01:00:48.264 --> 01:00:49.244
It's so clever, dude.

1062
01:00:49.884 --> 01:00:54.486
And what they're doing next, you talked about how this is just the beginning, right?

1063
01:00:54.786 --> 01:00:57.127
These pixels are about 100 microns in size.

1064
01:00:57.607 --> 01:01:01.708
They're trying to get it down to 20 micron.

1065
01:01:01.728 --> 01:01:04.289
20 micron pixels from 100 microns.

1066
01:01:04.629 --> 01:01:09.771
So now we'll be able to fit 10,000 pixels instead of the previous 378.

1067
01:01:09.931 --> 01:01:14.393
So now I'll be able to get as much fidelity as I do now.

1068
01:01:15.173 --> 01:01:17.954
Which is, again, now it's become an engineering problem.

1069
01:01:18.074 --> 01:01:18.394
Yes.

1070
01:01:18.414 --> 01:01:19.775
Which is much easier to deal with.

1071
01:01:19.995 --> 01:01:20.555
Yes, yes.

1072
01:01:20.575 --> 01:01:22.455
Like, I'll take an engineering problem all day.

1073
01:01:22.535 --> 01:01:23.536
Yes, yes.

1074
01:01:23.596 --> 01:01:24.876
20 microns, right?

1075
01:01:24.916 --> 01:01:29.118
That's about the size of the rods and cones themselves, right?

1076
01:01:29.138 --> 01:01:30.198
You're now starting to...

1077
01:01:30.558 --> 01:01:33.041
to get to the limit the biology was getting to.

1078
01:01:33.381 --> 01:01:35.163
Obviously, there's going to be challenges, right?

1079
01:01:35.223 --> 01:01:35.904
You can't...

1080
01:01:35.924 --> 01:01:43.532
Smaller pixel size means that the electric field penetration to the upper layers of that retina is going to be not as good.

1081
01:01:43.873 --> 01:01:46.676
So maybe you need these 3D electrode structures, right?

1082
01:01:46.696 --> 01:01:48.237
Where you have like pillars...

1083
01:01:48.918 --> 01:01:49.078
Right.

1084
01:01:49.118 --> 01:01:55.885
Where the pixel is like this and you've got a pillar that injects electricity into the upper layers of the retina.

1085
01:01:56.165 --> 01:02:03.552
But again, engineering problem, which which is infinitely more solvable than going from zero to one.

1086
01:02:04.673 --> 01:02:06.975
I mean, this thing is like decades in the making.

1087
01:02:07.576 --> 01:02:11.400
you know, but it's working and just, it's so clever.

1088
01:02:11.600 --> 01:02:12.060
I loved it.

1089
01:02:12.120 --> 01:02:12.901
Yeah.

1090
01:02:13.021 --> 01:02:16.605
No, that's really, it took a while to get access to the paper.

1091
01:02:16.625 --> 01:02:17.846
Oh yeah.

1092
01:02:18.026 --> 01:02:20.388
New England journal of medicine.

1093
01:02:20.408 --> 01:02:21.509
I don't know why you guys.

1094
01:02:21.630 --> 01:02:22.210
Yeah.

1095
01:02:22.270 --> 01:02:23.251
I don't, I don't want to pay.

1096
01:02:23.311 --> 01:02:24.412
Like, I think it was like a hundred.

1097
01:02:24.492 --> 01:02:24.993
I'll be honest.

1098
01:02:25.013 --> 01:02:26.294
I think it was like a hundred dollars.

1099
01:02:26.835 --> 01:02:29.717
So I asked a friend of mine who, who is a doctor.

1100
01:02:29.737 --> 01:02:29.837
Yeah.

1101
01:02:30.478 --> 01:02:31.298
Who does research.

1102
01:02:31.318 --> 01:02:32.619
Who has reason to access it.

1103
01:02:32.659 --> 01:02:34.139
Yeah, I need you to give me a PDF.

1104
01:02:34.219 --> 01:02:36.540
We're trying to give you guys some traffic here.

1105
01:02:36.820 --> 01:02:37.080
Okay?

1106
01:02:37.140 --> 01:02:38.340
We're trying to give some promo.

1107
01:02:38.840 --> 01:02:39.060
Yeah.

1108
01:02:39.480 --> 01:02:44.602
So, you know, free advertising on the First Principle pod.

1109
01:02:45.422 --> 01:02:46.742
We'll do deep dives on the stories.

1110
01:02:46.762 --> 01:02:47.943
Yeah, they're like, who are you?

1111
01:02:47.963 --> 01:02:49.983
And, like, honestly, give, like, good understanding.

1112
01:02:50.003 --> 01:02:57.085
Because a lot of times, again, like, for the average person, it's hard to really, like, parse and decipher what's actually happening in these stories.

1113
01:02:57.185 --> 01:02:58.306
And, like, what the implications are.

1114
01:02:58.326 --> 01:02:59.646
And what's, like, the real world...

1115
01:03:00.246 --> 01:03:05.147
you know, meaning and the real world meaning for this is like a big deal.

1116
01:03:05.247 --> 01:03:05.467
Yeah.

1117
01:03:05.527 --> 01:03:06.328
This is a big deal.

1118
01:03:06.348 --> 01:03:06.508
Right.

1119
01:03:06.828 --> 01:03:12.089
Like now imagine in the future right now, I'm thinking you can, you can bring AI into it.

1120
01:03:12.349 --> 01:03:12.689
Of course.

1121
01:03:12.769 --> 01:03:12.950
Right.

1122
01:03:13.270 --> 01:03:17.891
You could have like the, the, the Iron Man thing where like the thing automatically zooms.

1123
01:03:17.931 --> 01:03:18.111
Yeah.

1124
01:03:18.311 --> 01:03:20.031
So you don't have to press a button to zoom.

1125
01:03:20.132 --> 01:03:22.292
It'll just know what you're looking at and zoom in.

1126
01:03:23.272 --> 01:03:23.773
Like, yep.

1127
01:03:24.233 --> 01:03:28.614
No, I mean, this kind of like ties to the first story we had around the, how, how,

1128
01:03:29.274 --> 01:03:31.175
AI is now being used for cancer detection.

1129
01:03:31.315 --> 01:03:35.356
I mean, there's going to be clever ways in which to implement it in this context.

1130
01:03:35.676 --> 01:03:37.817
I mean, there's all this stuff.

1131
01:03:37.837 --> 01:03:42.099
I mean, these implants and these things, we've talked about Neuralink before on the pod.

1132
01:03:42.619 --> 01:03:45.540
Now we're having these retinal implants.

1133
01:03:45.820 --> 01:03:48.181
There's a bottleneck around the surgery process and stuff.

1134
01:03:48.201 --> 01:03:49.222
We're still meat sacks.

1135
01:03:49.242 --> 01:03:54.003
So that's still a bottleneck.

1136
01:03:54.404 --> 01:03:55.704
It's still a bottleneck, yeah.

1137
01:03:58.585 --> 01:03:59.706
And it's going to be for a while.

1138
01:03:59.746 --> 01:04:00.247
For a while.

1139
01:04:00.287 --> 01:04:03.609
But we're getting into real interesting places.

1140
01:04:03.889 --> 01:04:09.693
I mean, as someone who already has bad vision, you know, my wife has bad vision too.

1141
01:04:10.014 --> 01:04:12.776
She actually has – I didn't do LASIK or anything.

1142
01:04:12.836 --> 01:04:14.317
I just wear contacts and glasses.

1143
01:04:15.017 --> 01:04:19.480
But her vision was so bad that at the time, like, LASIK wasn't good enough for her eyes.

1144
01:04:19.801 --> 01:04:22.162
So she got the implanted contact lens.

1145
01:04:22.503 --> 01:04:23.143
Oh.

1146
01:04:23.243 --> 01:04:24.564
Like in her eyes.

1147
01:04:24.824 --> 01:04:25.205
Wow.

1148
01:04:25.385 --> 01:04:26.806
Which is like – and I was like –

1149
01:04:27.166 --> 01:04:28.988
That's crazy that you can do that.

1150
01:04:29.068 --> 01:04:31.811
And this was like 15 years, you know, whatever ago.

1151
01:04:31.831 --> 01:04:32.712
This was a while ago.

1152
01:04:33.733 --> 01:04:37.216
And at the time, it was state of the art and this and that and yada yada.

1153
01:04:37.657 --> 01:04:38.478
But, I mean, this is like...

1154
01:04:39.396 --> 01:04:46.440
orders of magnitude more difficult than just embedding a contact lens in the eye.

1155
01:04:47.020 --> 01:04:48.821
Fascinating, fascinating, fascinating.

1156
01:04:48.841 --> 01:04:50.181
Yeah, I thought that was such a cool story.

1157
01:04:50.241 --> 01:04:51.062
That's a great story.

1158
01:04:51.462 --> 01:04:53.223
That is a great, great story.

1159
01:04:54.163 --> 01:04:58.606
It gives me hope that if I start getting blind later in life, there's still a chance.

1160
01:04:58.626 --> 01:04:59.586
Yeah, there's still a chance.

1161
01:04:59.606 --> 01:05:00.066
There's still a chance.

1162
01:05:00.086 --> 01:05:01.187
So you're saying there's a chance.

1163
01:05:01.207 --> 01:05:02.588
You're saying there's a chance.

1164
01:05:03.228 --> 01:05:07.170
We're going to end with our story number three, which is our story about...

1165
01:05:08.766 --> 01:05:13.408
or related to at least unidentified anomalous phenomenon, which is more commonly known as UFOs.

1166
01:05:13.788 --> 01:05:34.056
And before we kind of get into the two papers themselves, I kind of wanted to set the table a little bit because this is a subject and an issue that is extremely popular, not only in the US, but globally, and has in recent years gotten a boost because of a variety of

1167
01:05:35.772 --> 01:05:45.020
events that have been happening, particularly within the apparatus of the U.S. government, intelligence, and military-industrial complex structures.

1168
01:05:45.920 --> 01:05:48.762
And, you know, this is a subject I've talked to you about a lot.

1169
01:05:48.783 --> 01:05:50.144
Yeah, and you're very passionate about it.

1170
01:05:50.584 --> 01:05:53.206
It's something that I have interest in.

1171
01:05:53.386 --> 01:05:57.950
I work at a nonprofit that's working in this space, the Disclosure Foundation, that's

1172
01:05:58.370 --> 01:06:01.031
And one of the things that people always get caught up on is origin.

1173
01:06:01.692 --> 01:06:08.215
And what I would ask is that we sort of keep origin to the side for a second in this conversation.

1174
01:06:08.895 --> 01:06:12.417
And just talk about what is without trying to ascribe an origin to it.

1175
01:06:12.437 --> 01:06:17.980
Because these are all things that are happening regardless of what you want to ascribe the origin to.

1176
01:06:18.840 --> 01:06:22.502
And it's interesting regardless of what the origin is.

1177
01:06:23.613 --> 01:06:34.296
And so I'm going to do a quick retrospective on kind of the recent history of the UAP subject that is the context in which these two papers are coming out.

1178
01:06:35.336 --> 01:06:46.958
And so for many people, they're probably familiar with the 2017 New York Times article that came out that basically talked about the fact that there's these U.S. government programs that are studying UFOs that have existed for some time.

1179
01:06:48.539 --> 01:06:50.959
And that was sort of the catalyst for the modern era.

1180
01:06:53.740 --> 01:06:57.762
Prior to that, that is outside of the scope of this episode.

1181
01:06:57.782 --> 01:06:58.002
Yeah.

1182
01:06:58.522 --> 01:07:05.886
And so we're not going to touch on sort of the, you know, the what I call the legacy era of the subject.

1183
01:07:07.026 --> 01:07:10.848
But after that New York Times article came out in 2017, it contained –

1184
01:07:12.449 --> 01:07:13.409
three videos.

1185
01:07:13.949 --> 01:07:20.431
This was the first time we were getting videos with chain of custody from the U S government that were showing these anomalous objects.

1186
01:07:20.651 --> 01:07:20.811
Right.

1187
01:07:20.831 --> 01:07:21.091
Okay.

1188
01:07:21.551 --> 01:07:31.273
But it was still only put out by the New York times, which people will have different and differing opinions about the quality of that institution as a newspaper.

1189
01:07:32.214 --> 01:07:33.794
That's changed since that time period.

1190
01:07:34.274 --> 01:07:39.535
But the Pentagon actually confirmed that the, what are colloquially known as the gimbal,

1191
01:07:42.596 --> 01:07:45.360
are legitimate videos that have chain of custody to the US government.

1192
01:07:45.580 --> 01:07:48.163
That happened between 2017 and 2019.

1193
01:07:49.425 --> 01:07:54.651
What sort of proceeded from there was an escalation of the...

1194
01:07:56.085 --> 01:07:59.048
issue from a national security perspective.

1195
01:07:59.648 --> 01:08:22.087
So in 2020, the then Senate Intel minority leader, Marco Rubio, who is now the national security advisor in the Trump administration, was one of several senators who called for there to be an unclassified report to be created by the Department of Defense around the

1196
01:08:23.068 --> 01:08:24.509
What are, what are these UAP?

1197
01:08:24.629 --> 01:08:26.431
Where is this unidentified anomalous phenomena?

1198
01:08:26.871 --> 01:08:50.947
In 2021, the office of the director of national intelligence delivered that report in which they stated that there were 143 unexplained cases that, you know, again, this was largely driven by Navy and other pilots basically saying, look, this is a safety of flight issue and we need to like be able to report and address this just for our day to day operations.

1199
01:08:51.047 --> 01:08:51.228
Right.

1200
01:08:51.488 --> 01:08:51.688
Again,

1201
01:08:52.048 --> 01:08:54.210
independent of origin or any of these other issues.

1202
01:08:55.351 --> 01:09:10.445
The DOD then created, you know, this UAP task force, which was this sort of temporary task force that went out to both the intelligence agencies and the military sort of apparatus to try to understand what,

1203
01:09:12.606 --> 01:09:13.687
Are you all seeing this?

1204
01:09:14.187 --> 01:09:15.567
Where are you seeing this stuff?

1205
01:09:16.448 --> 01:09:17.568
What is going on here?

1206
01:09:18.228 --> 01:09:29.132
That eventually now rolled into what we now know as this permanent Pentagon office that's called the All Domain Anomaly Resolution Office.

1207
01:09:29.752 --> 01:09:31.353
That was established in 2022.

1208
01:09:32.794 --> 01:09:39.057
And that office was congressionally mandated to be created after that initial report came out.

1209
01:09:39.397 --> 01:09:52.785
Because Congress is basically alleging that the executive branch, which runs the intelligence agencies and the military, are not providing them the information about what's happening as it relates to this subject specifically.

1210
01:09:53.185 --> 01:09:57.728
And it's important to note that the term UAP is a specific term of art.

1211
01:09:58.408 --> 01:10:03.030
that does not include any other form of unmanned aerial systems.

1212
01:10:03.590 --> 01:10:08.432
So this is drones, this is undersea, unmanned.

1213
01:10:08.492 --> 01:10:14.514
Those are all discrete and defined, and there's processes we have to deal with that.

1214
01:10:14.554 --> 01:10:20.717
The reason why this became an issue within the government is that these platforms were exhibiting

1215
01:10:22.097 --> 01:10:31.826
like characteristics or behaviors that were beyond the scope of what was deemed as current or even next generation technologies.

1216
01:10:32.346 --> 01:10:43.777
This has then been followed up by multiple hearings within the House and the Senate on this issue, and it's culminated with legislation that part of which was passed previously,

1217
01:10:44.301 --> 01:10:55.364
that was proposed initially in 2023, but then passed in 2024, that mandated that executive agencies release information around the history of the subject.

1218
01:10:55.704 --> 01:11:12.429
I mean, there's literally a colloquy between Senator Schumer, who was the then Senate Majority Leader for the Democrats, and Senator Mike Rounds, who's a Republican, where they were sort of going back and forth, identifying that this is something that

1219
01:11:13.519 --> 01:11:21.109
has been an ongoing issue that Congress has been going after and the executive branch has been unwilling to provide information on.

1220
01:11:21.589 --> 01:11:23.632
So this is sort of a live issue.

1221
01:11:23.672 --> 01:11:26.255
NASA had a UAP independent study group.

1222
01:11:26.536 --> 01:11:28.578
They put out their own report around this issue.

1223
01:11:28.598 --> 01:11:28.658
Um,

1224
01:11:30.120 --> 01:11:37.907
They are obviously struggling for funding, so it's not something that they're spending money on because they're trying to just survive as it is.

1225
01:11:38.427 --> 01:11:41.670
There's been multiple whistleblower protections that have been passed.

1226
01:11:41.950 --> 01:11:45.313
There's been defunding of UAP programs for the intelligence community.

1227
01:11:45.333 --> 01:11:46.514
I mean, there's this long list of...

1228
01:11:47.955 --> 01:11:52.141
of indirect evidence around this subject.

1229
01:11:52.522 --> 01:11:55.406
But there's been a dearth of scientific research.

1230
01:11:55.707 --> 01:11:55.947
Yes.

1231
01:11:56.267 --> 01:12:02.857
Really the main primary thing that has existed in this space, particularly in the last couple of years, have been research papers...

1232
01:12:03.638 --> 01:12:12.425
that have been trying to define how to study the subject because of the reproducibility problem and the stigma, both of which are big issues.

1233
01:12:13.005 --> 01:12:22.773
And the reference point that has been addressed is like multi-messenger astronomy studies has been the reference point for how to try to deal with this problem.

1234
01:12:23.353 --> 01:12:32.300
But there are two papers that have just come out that are a first of its kind in terms of having actual data

1235
01:12:33.777 --> 01:12:41.447
that again can be potentially either reproduced or replicated in other arenas from the Palomar Sky Survey.

1236
01:12:41.987 --> 01:12:48.676
And we're going to kind of go in from a first principles perspective to understand what these papers actually were talking about.

1237
01:12:49.036 --> 01:12:49.277
Yeah.

1238
01:12:49.857 --> 01:12:50.058
Yeah.

1239
01:12:50.218 --> 01:12:51.279
So there's a lead author.

1240
01:12:52.221 --> 01:12:54.842
Dr. Beatrice Villarreal.

1241
01:12:55.082 --> 01:12:55.862
Villarreal, yes.

1242
01:12:56.122 --> 01:12:56.743
Villarreal.

1243
01:12:56.883 --> 01:12:57.043
Yes.

1244
01:12:57.483 --> 01:12:58.403
I hope I'm saying that right.

1245
01:12:58.483 --> 01:12:58.663
Yes.

1246
01:12:58.943 --> 01:13:06.786
So the lead author, Dr. Beatrice Villarreal, she actually covered this issue before in her 2021 paper.

1247
01:13:06.906 --> 01:13:07.586
Yes.

1248
01:13:07.626 --> 01:13:09.627
That was published in Nature Scientific Reports.

1249
01:13:09.727 --> 01:13:14.369
It was on nine simultaneous transients occurred in April 1950.

1250
01:13:14.389 --> 01:13:14.489
Yeah.

1251
01:13:16.049 --> 01:13:21.312
And it's important to note that research on this has been done.

1252
01:13:21.332 --> 01:13:21.953
Has been done.

1253
01:13:22.093 --> 01:13:27.896
And this is kind of a follow-up paper, but it's, I believe, a lot more thorough.

1254
01:13:28.397 --> 01:13:37.962
And I just thought that the way that the research was conducted, the science sort of speaks for itself, and then the interpretations come afterwards.

1255
01:13:38.082 --> 01:13:38.283
Sure.

1256
01:13:38.323 --> 01:13:38.523
But...

1257
01:13:39.310 --> 01:13:44.713
You know, on this podcast, we always focus on from first principles, what is the science that you did?

1258
01:13:45.354 --> 01:13:51.857
And does it warrant the conclusions that you put out?

1259
01:13:52.078 --> 01:13:52.258
Yes.

1260
01:13:52.478 --> 01:13:52.678
Right?

1261
01:13:53.038 --> 01:13:55.099
So we're going to start with these papers.

1262
01:13:55.980 --> 01:13:59.542
The papers focus on the Palomar Observatory Sky Survey.

1263
01:13:59.962 --> 01:14:00.222
Okay?

1264
01:14:01.023 --> 01:14:02.864
This is POSS1.

1265
01:14:04.025 --> 01:14:07.767
It was a sky survey that was taken up by the Samuel Ostgen Telescope.

1266
01:14:09.148 --> 01:14:13.070
which is a camera that is mounted on the Samuel Oshin Telescope at Palomar Observatory.

1267
01:14:13.410 --> 01:14:15.331
There you can see the telescope itself.

1268
01:14:15.391 --> 01:14:16.551
It's got a Schmidt camera.

1269
01:14:16.951 --> 01:14:19.932
And this thing is specialized for wide field photography.

1270
01:14:20.353 --> 01:14:22.393
These cameras are actually pretty insane because...

1271
01:14:23.234 --> 01:14:23.454
So...

1272
01:14:24.645 --> 01:14:32.551
That camera, the camera that is attached to it, has a field of view of 6.6 degrees by 6.6 degrees.

1273
01:14:32.831 --> 01:14:35.573
Okay, so 6.6 degrees by 6.6 degrees.

1274
01:14:35.973 --> 01:14:39.496
The full moon, for comparison, is half a degree.

1275
01:14:39.836 --> 01:14:45.580
It's about like if you were to put your thumb right in front of your face, the size of your thumb is about half a degree.

1276
01:14:46.120 --> 01:14:47.882
This is 13 full moons by 13 full moons.

1277
01:14:47.922 --> 01:14:49.002
So it's 170 full moons.

1278
01:14:49.263 --> 01:14:49.423
Mm-hmm.

1279
01:14:53.147 --> 01:14:56.030
Big is the field of view of this camera, right?

1280
01:14:56.510 --> 01:15:02.195
And because it's so big, there's a specialized camera called a Schmidt camera.

1281
01:15:02.515 --> 01:15:05.258
And what it does is you've got a primary mirror, right?

1282
01:15:05.298 --> 01:15:07.159
The mirror is about 1.2 meters.

1283
01:15:07.680 --> 01:15:09.822
That mirror then goes into the camera.

1284
01:15:10.826 --> 01:15:14.907
You can imagine the light is getting warped, right?

1285
01:15:15.307 --> 01:15:17.188
As the light comes in, it gets warped.

1286
01:15:17.288 --> 01:15:21.629
And so the edges of your field of view are going to get distorted.

1287
01:15:21.929 --> 01:15:22.089
Yep.

1288
01:15:22.489 --> 01:15:22.649
Yep.

1289
01:15:23.169 --> 01:15:27.950
This specific camera is specialized for that kind of wide field photography.

1290
01:15:28.311 --> 01:15:30.111
So it corrects for that distortion.

1291
01:15:30.471 --> 01:15:33.312
And even then, the focal plane...

1292
01:15:34.132 --> 01:15:35.353
is actually not flat.

1293
01:15:35.694 --> 01:15:44.382
You know how in like a camera, your CCD is basically a flat piece of equipment because all of your light is coming onto a flat piece of equipment.

1294
01:15:45.042 --> 01:15:50.087
Here, because it's such a large field of view and it's so big, the focal...

1295
01:15:51.039 --> 01:15:52.740
the focal plane is actually curved.

1296
01:15:53.221 --> 01:16:04.350
So in order to get a photograph, your photograph, the plate in this case, the photographic plate, and these are all photographic plates that are used, also has to be curved.

1297
01:16:04.710 --> 01:16:09.033
And so you got to like warp it before you slide it in there and get this photograph.

1298
01:16:09.133 --> 01:16:10.154
I thought that was really cool.

1299
01:16:10.194 --> 01:16:10.615
That's interesting.

1300
01:16:10.635 --> 01:16:18.021
Like the amount of stuff that needed to happen in the 1950s in order to get these kinds of large field of view photographs, right?

1301
01:16:18.181 --> 01:16:19.322
Given the technology ahead of time.

1302
01:16:19.342 --> 01:16:19.822
Now, this is pretty...

1303
01:16:19.882 --> 01:16:27.553
Purely photographic, which means that the other cool thing is that the telescope that we just saw...

1304
01:16:28.698 --> 01:16:31.260
That telescope is an equatorial mount.

1305
01:16:31.320 --> 01:16:35.564
And what that means is that one of the axes is parallel to the Earth's axis of rotation.

1306
01:16:35.984 --> 01:16:42.811
And so all you have to do is rotate along that axis, and it'll follow the stars as the Earth rotates underneath.

1307
01:16:43.251 --> 01:16:46.674
The stars are all going to move in a circle around the North Star.

1308
01:16:47.034 --> 01:16:55.782
And so in order to track a single spot in the sky, the telescope has to rotate with the Earth so that it points in one direction directly.

1309
01:16:55.942 --> 01:16:57.203
in the celestial sphere.

1310
01:16:57.623 --> 01:17:16.450
And in order to actually do that back then, they had a manual astronomer that would look through a 10 inch smaller telescope to a guide star and then, and then control the telescope manually to keep that guide star in the exact same position.

1311
01:17:16.490 --> 01:17:19.472
So there was a lot of like manual stuff going on back in the day.

1312
01:17:19.492 --> 01:17:22.573
And I think this is really important to see, to, to understand that,

1313
01:17:23.293 --> 01:17:25.175
how the data was collected, right?

1314
01:17:25.235 --> 01:17:44.692
It's not with this modern technology where now everything is automated and nothing is actually in equatorial because it's easier to make bulkier telescopes in the azimuthal mount where I can control both axes with a computer and I just have to do the calculation and control both axes and I'm fine.

1315
01:17:45.533 --> 01:17:48.896
So this guy's survey produced a thousand photographic plates

1316
01:17:50.065 --> 01:17:50.805
About 1,000.

1317
01:17:51.645 --> 01:17:54.726
One in red, one in blue, so then you'd have about 2,000.

1318
01:17:55.406 --> 01:17:57.867
This is an example of one of the photographic plates.

1319
01:17:58.787 --> 01:18:05.089
It's a one degree by one degree patch of the plate that is targeting Pleiades, the seven sisters.

1320
01:18:05.769 --> 01:18:09.630
And the haziness that you see, that's actually not a defect.

1321
01:18:10.090 --> 01:18:16.712
Those are actual clouds around those stars, those dust clouds around those stars that are creating that.

1322
01:18:16.732 --> 01:18:17.052
Right.

1323
01:18:17.092 --> 01:18:22.894
It's a beautiful, beautiful example of one of the photographic plates that was actually taken during that time.

1324
01:18:23.554 --> 01:18:27.655
The typical exposure was about 45 to 50 minutes.

1325
01:18:27.815 --> 01:18:30.596
What we're focusing on here are the red-sensitive plates.

1326
01:18:30.956 --> 01:18:33.097
They used to take red-sensitive and blue-sensitive.

1327
01:18:33.137 --> 01:18:37.478
Here we're focusing on the red-sensitive because that's what was used in the study here.

1328
01:18:37.578 --> 01:18:42.299
And the limiting magnitude is 22 in terms of apparent brightness.

1329
01:18:42.379 --> 01:18:44.460
The higher the magnitude, the dimmer.

1330
01:18:45.040 --> 01:18:45.720
the thingy is.

1331
01:18:46.100 --> 01:18:50.442
And this is about 1 million times fainter than what the human eye can see.

1332
01:18:50.502 --> 01:18:50.782
Okay.

1333
01:18:50.822 --> 01:18:55.364
So that's what's limiting the faintest object that this telescope can see.

1334
01:18:55.384 --> 01:18:55.784
Can see.

1335
01:18:56.024 --> 01:18:56.264
Okay.

1336
01:18:56.304 --> 01:18:57.745
That's like the lower bound.

1337
01:18:57.785 --> 01:18:58.965
Yes, that is the lower bound.

1338
01:18:59.005 --> 01:19:00.646
So that's the telescope that we're using.

1339
01:19:00.666 --> 01:19:00.966
Okay.

1340
01:19:00.986 --> 01:19:02.587
And that's how we're getting these images.

1341
01:19:02.867 --> 01:19:03.027
Okay.

1342
01:19:03.047 --> 01:19:05.988
Now, the photographic plates.

1343
01:19:06.008 --> 01:19:06.588
Okay.

1344
01:19:06.840 --> 01:19:08.201
We got to talk about the photographic plates.

1345
01:19:08.221 --> 01:19:09.161
We got to talk about the plates.

1346
01:19:09.381 --> 01:19:10.021
What does that mean?

1347
01:19:10.361 --> 01:19:10.581
Yes.

1348
01:19:10.601 --> 01:19:13.582
So photographic plates are 14 by 14 inches.

1349
01:19:14.203 --> 01:19:19.965
And these 14 by 14 inches are placed at the back of the camera where all of the light is coming in.

1350
01:19:20.165 --> 01:19:21.685
It's one millimeter thick.

1351
01:19:22.126 --> 01:19:24.306
These are incredibly thin for how big they are.

1352
01:19:24.326 --> 01:19:27.688
They're like the size of a record, you know, a vinyl record.

1353
01:19:27.708 --> 01:19:28.428
Yeah.

1354
01:19:30.161 --> 01:19:34.203
They've got a photographic emulsion on them, okay?

1355
01:19:34.303 --> 01:19:37.124
A light-sensitive silver halide crystal layer.

1356
01:19:37.784 --> 01:19:43.886
Usually it's silver bromide, and that's suspended in gelatin, okay?

1357
01:19:43.906 --> 01:19:45.907
Bromium is the halogen that they're using here.

1358
01:19:46.187 --> 01:19:50.469
Halogen is like that column on the periodic table that's second to last on the right.

1359
01:19:50.489 --> 01:19:50.849
Okay.

1360
01:19:51.089 --> 01:19:59.072
These plates are bent, as I was saying, so that they can match the focal plane, which I thought was just really cool to think about.

1361
01:19:59.172 --> 01:20:06.656
And then the process is a photon comes in from your telescope and it liberates an electron in the silver bromide.

1362
01:20:06.976 --> 01:20:12.058
And that electron, that liberation of the electron forms a latent image spec.

1363
01:20:12.558 --> 01:20:13.739
You've got this free...

1364
01:20:14.720 --> 01:20:17.484
silver atom, okay, that creates a silver atom cluster.

1365
01:20:17.885 --> 01:20:29.721
Then during development, that latent image spec catalyzes, and the silver bromide that's around there goes and turns into black metallic silver, okay?

1366
01:20:30.540 --> 01:20:36.502
And then finally you have a washer, a kind of fixer that removes any unexposed silver bromide.

1367
01:20:36.802 --> 01:20:39.703
This is like the old school, you know, where you like wash the thingy.

1368
01:20:40.083 --> 01:20:43.044
And then that removes any silver bromide.

1369
01:20:43.524 --> 01:20:46.225
And what you're left with is a permanent negative image.

1370
01:20:46.485 --> 01:20:55.969
And that's why in the previous image that you saw, if you actually go back to photo two, you'll see in the previous image, the dark spots are actually where the light is.

1371
01:20:56.109 --> 01:20:56.649
Okay?

1372
01:20:56.669 --> 01:20:56.909
Right.

1373
01:20:56.969 --> 01:20:57.169
Right.

1374
01:20:57.489 --> 01:21:00.017
Because that's where the silver is getting deposited.

1375
01:21:00.078 --> 01:21:00.379
Yes.

1376
01:21:00.399 --> 01:21:00.539
Okay?

1377
01:21:00.780 --> 01:21:02.505
So you get this negative image.

1378
01:21:02.526 --> 01:21:02.766
Yeah.

1379
01:21:03.092 --> 01:21:03.552
Right.

1380
01:21:03.772 --> 01:21:06.253
Of the dark is actually where the light is coming in.

1381
01:21:06.593 --> 01:21:06.773
Got it.

1382
01:21:06.793 --> 01:21:07.013
Okay?

1383
01:21:07.233 --> 01:21:10.974
And the white is where there was no photo reaction that happened.

1384
01:21:11.314 --> 01:21:11.534
Okay?

1385
01:21:11.954 --> 01:21:16.635
And so that's how we actually get these photographic plates.

1386
01:21:16.655 --> 01:21:17.655
That's what a photographic plate is.

1387
01:21:17.675 --> 01:21:19.955
That's what a photographic plate is.

1388
01:21:20.295 --> 01:21:24.596
And that is what the chemical reaction of what a photographic plate means.

1389
01:21:25.136 --> 01:21:25.376
Okay?

1390
01:21:26.117 --> 01:21:27.537
Now, what this...

1391
01:21:28.457 --> 01:21:40.389
What this sky survey is doing and the data that this particular group used is the digitized version of those photographic plates.

1392
01:21:40.650 --> 01:21:50.520
So the digitized version in the 1980s and 1990s, we did a digitization process where those glass plates used micro densitometers for...

1393
01:21:51.020 --> 01:21:54.682
that basically went through and digitized these photographic plates.

1394
01:21:54.702 --> 01:22:03.528
You've got a light beam that measures the opacity of the photographic plate, and that converts that opacity into a digital pixel value.

1395
01:22:03.888 --> 01:22:08.390
So the more opaque it is, that means that there was more dark, right?

1396
01:22:08.410 --> 01:22:09.511
There was more darkness.

1397
01:22:09.911 --> 01:22:16.755
And then if there's more darkness, that means that there's more light over there.

1398
01:22:16.876 --> 01:22:19.097
And so now you've converted from...

1399
01:22:21.014 --> 01:22:25.055
Photographic plates to a digital sort of pixel values.

1400
01:22:25.095 --> 01:22:25.375
Yes.

1401
01:22:25.415 --> 01:22:25.635
Right?

1402
01:22:25.875 --> 01:22:26.055
Yes.

1403
01:22:26.155 --> 01:22:30.536
And there were two digitization processes that happened.

1404
01:22:30.556 --> 01:22:30.816
Okay.

1405
01:22:30.836 --> 01:22:37.017
There was the digital sky survey that was done with a pixel size of 25 microns.

1406
01:22:37.218 --> 01:22:37.398
Okay.

1407
01:22:37.418 --> 01:22:40.138
And then there was a super cosmos sky survey.

1408
01:22:40.158 --> 01:22:40.298
Yep.

1409
01:22:40.638 --> 01:22:44.059
And that was done at the Royal Observatory of Edinburgh.

1410
01:22:46.193 --> 01:22:51.978
With a higher resolution of 0.7 arc seconds per pixel.

1411
01:22:52.339 --> 01:22:52.599
Got it.

1412
01:22:52.819 --> 01:22:53.079
Okay.

1413
01:22:53.380 --> 01:22:56.282
So the first one was 1.7 arc seconds per pixel.

1414
01:22:56.603 --> 01:22:58.805
And then this one's 0.7 arc seconds.

1415
01:22:58.825 --> 01:23:02.728
So it's just like more grainier detail.

1416
01:23:02.768 --> 01:23:03.089
Yeah.

1417
01:23:03.169 --> 01:23:03.389
You know?

1418
01:23:03.769 --> 01:23:05.511
Sort of a thing about.

1419
01:23:05.551 --> 01:23:06.672
Slightly higher fidelity.

1420
01:23:06.712 --> 01:23:06.932
Yes.

1421
01:23:07.192 --> 01:23:07.653
Exactly.

1422
01:23:08.033 --> 01:23:09.835
And now you've got a vetting step, right?

1423
01:23:09.875 --> 01:23:10.856
Where now...

1424
01:23:12.965 --> 01:23:20.849
We can create computer algorithms that look at these digital photographic versions and then try to find transients.

1425
01:23:21.209 --> 01:23:30.033
Little spots in the sky that aren't there in current photographs of that patch of the sky.

1426
01:23:30.093 --> 01:23:31.033
But that were there.

1427
01:23:31.093 --> 01:23:32.834
That were there back then.

1428
01:23:33.154 --> 01:23:39.297
And what we can also do is we can rule out any scanning artifacts because we've got two different digital versions that...

1429
01:23:40.250 --> 01:23:41.511
Of the photographic plate.

1430
01:23:41.851 --> 01:23:47.636
So if something happened in the scanning where I got a little spec, that shouldn't be there in both at the same spot.

1431
01:23:47.716 --> 01:23:48.357
Yeah, yeah, right.

1432
01:23:48.377 --> 01:23:49.758
The probability of that is insane.

1433
01:23:49.778 --> 01:23:50.859
It's insanely high.

1434
01:23:51.019 --> 01:23:55.883
The point is the digital scans, there's two versions.

1435
01:23:57.204 --> 01:24:06.372
And so if someone was asleep at the wheel when they scanned version one, it should not be there in version two at the exact same spot.

1436
01:24:07.761 --> 01:24:10.624
In such a precise way.

1437
01:24:10.944 --> 01:24:13.045
And so what you're sort of...

1438
01:24:13.085 --> 01:24:16.388
There's a layer of validation.

1439
01:24:16.408 --> 01:24:17.889
Yeah, there's like a redundancy here.

1440
01:24:17.909 --> 01:24:18.689
A redundancy here.

1441
01:24:18.890 --> 01:24:22.132
Because we have two digital... Yeah, we have two digital versions of this.

1442
01:24:22.352 --> 01:24:22.592
Okay.

1443
01:24:22.632 --> 01:24:23.113
And so...

1444
01:24:24.234 --> 01:24:30.599
Now we get into the Vasco project, right, which is the vanishing and appearing sources during a century of observations.

1445
01:24:30.939 --> 01:24:33.621
This is led by Beatrice Villarreal.

1446
01:24:34.962 --> 01:24:44.089
And the point of this project is to have a systematic comparison between the digitized versions and modern sky surveys.

1447
01:24:44.129 --> 01:24:50.434
You look at what the Palomar Observatory saw back then, and you look at what we're looking at right now.

1448
01:24:50.914 --> 01:24:57.460
And you've got an automated software that flags these vanishing, the present then, missing now, right?

1449
01:24:57.660 --> 01:24:58.822
And there's also appearing.

1450
01:24:59.282 --> 01:25:03.286
So not present then, but appearing now, right?

1451
01:25:03.306 --> 01:25:05.628
You've got an automated software that flags these things.

1452
01:25:06.168 --> 01:25:07.790
And initially, you've got a catalog.

1453
01:25:08.922 --> 01:25:11.765
of 100,000 transient candidates.

1454
01:25:12.126 --> 01:25:18.213
And here we're seeing two different images, one from POS1, which is the first sky survey, and POS2.

1455
01:25:18.573 --> 01:25:20.055
And the circles represent...

1456
01:25:20.911 --> 01:25:26.698
two little spots that you can see are exactly not there in the second one, right?

1457
01:25:26.758 --> 01:25:31.063
It's like pretty obvious that those two dots were there before and now they're not there now, right?

1458
01:25:31.403 --> 01:25:32.865
And so that's a transient.

1459
01:25:32.885 --> 01:25:37.730
You've got a point-like source that's there in some plates, but in the later plates, it's not there.

1460
01:25:38.251 --> 01:25:40.794
And the idea is, you know, we have two points in time.

1461
01:25:41.875 --> 01:25:44.336
And we should be seeing identical things in both those points in time.

1462
01:25:44.476 --> 01:25:44.716
Yes.

1463
01:25:45.036 --> 01:25:46.416
If it's a star, then yeah.

1464
01:25:46.617 --> 01:25:49.197
Because of... And it's not identical.

1465
01:25:49.217 --> 01:25:50.038
It's not identical.

1466
01:25:50.598 --> 01:25:50.798
Right?

1467
01:25:51.358 --> 01:25:57.080
So, people are going to be skeptical about this, right?

1468
01:25:57.100 --> 01:25:57.260
Right?

1469
01:25:58.502 --> 01:26:01.124
For example, it could just be plate defects.

1470
01:26:01.764 --> 01:26:01.964
Sure.

1471
01:26:02.405 --> 01:26:02.625
Right?

1472
01:26:02.845 --> 01:26:12.652
Both this point of source of light or whatever, this thing that we're seeing in the photographic plate, it's there in both the digital versions.

1473
01:26:12.952 --> 01:26:14.814
Well, that means it could just be in the plate.

1474
01:26:15.538 --> 01:26:16.118
itself.

1475
01:26:16.638 --> 01:26:19.059
What is a hallmark of something that is in the plate?

1476
01:26:19.119 --> 01:26:23.220
And this is something that comes from Nigel Hambly at the University of Edinburgh.

1477
01:26:23.260 --> 01:26:24.081
He's one of the critics.

1478
01:26:24.481 --> 01:26:30.402
He looks at something called the full width at half maximum of these points of light.

1479
01:26:30.902 --> 01:26:34.524
So the full width at half max is basically, imagine you've got a Gaussian.

1480
01:26:36.104 --> 01:26:38.885
What you're looking for is what is the width of that Gaussian?

1481
01:26:39.585 --> 01:26:43.508
When the height is half the peak height, okay?

1482
01:26:44.208 --> 01:26:50.413
And what this tells you is something about the noise in the data gathering process.

1483
01:26:50.453 --> 01:26:56.056
Now, I told you this telescope, right?

1484
01:26:56.397 --> 01:26:57.778
It's traversing the night sky.

1485
01:26:58.018 --> 01:26:58.218
Yes.

1486
01:26:59.019 --> 01:27:06.167
and it's being controlled by a human being, well, there's going to be a little bit of noise in how I control that telescope as it tracks, right?

1487
01:27:06.428 --> 01:27:10.252
So even a stationary point of light is going to have a little bit of jitter

1488
01:27:11.475 --> 01:27:16.077
As I move this telescope, the telescope is going to not totally track it completely.

1489
01:27:16.257 --> 01:27:22.960
And even a point of light like a star that is definitely stationary over the 50 minutes at least.

1490
01:27:23.000 --> 01:27:26.902
It might be moving over several years, but over 50 minutes it's stationary.

1491
01:27:27.182 --> 01:27:30.744
That's going to have a little spread in my photographic plate.

1492
01:27:31.104 --> 01:27:35.706
And what he noticed was these transients, the ones that sort of appear and disappear,

1493
01:27:36.772 --> 01:27:42.896
They've got a spread that's way smaller than the stars.

1494
01:27:43.837 --> 01:27:44.157
Okay?

1495
01:27:44.637 --> 01:27:44.997
Yeah.

1496
01:27:45.177 --> 01:27:48.400
And that, to him, was a telltale sign that this is a defect.

1497
01:27:48.940 --> 01:27:55.544
Because if it was a light source, then the light source should have the same sort of error that the stars do.

1498
01:27:55.724 --> 01:27:56.445
Because they're all light.

1499
01:27:56.465 --> 01:27:56.725
Yeah.

1500
01:27:56.865 --> 01:27:57.065
Right?

1501
01:27:57.185 --> 01:27:57.786
Does that make sense?

1502
01:27:57.806 --> 01:27:58.066
It does.

1503
01:27:58.126 --> 01:27:58.346
It does.

1504
01:27:58.386 --> 01:27:58.546
Right?

1505
01:27:58.586 --> 01:28:03.951
Like all of the other things have the same sort of spread because it's from the same jitter of the telescope.

1506
01:28:03.971 --> 01:28:05.171
It's uniform across the system.

1507
01:28:05.352 --> 01:28:05.592
Right?

1508
01:28:05.732 --> 01:28:06.933
Yeah, yeah, yeah.

1509
01:28:06.953 --> 01:28:07.613
It's a key thing.

1510
01:28:07.653 --> 01:28:07.954
Yeah, yeah.

1511
01:28:08.114 --> 01:28:17.201
Now, the contention is, the sort of counter argument is, well, if it's truly transient, meaning like it only like lit up.

1512
01:28:17.870 --> 01:28:29.256
for a minute or a fraction of a second, and it had a bright signature, then only one part of that 50-minute exposure is going to have that light come on, and then it's going to go away.

1513
01:28:29.556 --> 01:28:38.441
So then, of course, the full width at half max is going to be way shorter, because it's not going to have time to accumulate that noise, right?

1514
01:28:38.761 --> 01:28:41.683
But now we're here, we're at a degeneracy problem, right?

1515
01:28:41.763 --> 01:28:45.865
Because a genuinely short-lived astronomical event

1516
01:28:46.920 --> 01:28:52.202
something that happens for a fraction of a second, is going to look identical to a plate artifact.

1517
01:28:52.862 --> 01:28:54.043
Got it.

1518
01:28:54.883 --> 01:29:00.705
It's like you're looking for a short-lived astronomical event, but that's going to look exact.

1519
01:29:01.746 --> 01:29:10.489
How do you distinguish between what is being argued as a plate artifact versus a true short-lived transient event?

1520
01:29:10.769 --> 01:29:13.850
Because it's hard to distinguish between those two.

1521
01:29:14.050 --> 01:29:15.310
Yeah, yeah, yeah.

1522
01:29:15.370 --> 01:29:22.432
The measure that we're using is degenerate to discerning between one or the other.

1523
01:29:22.452 --> 01:29:24.253
Does that make sense?

1524
01:29:24.433 --> 01:29:25.293
It's one of the two.

1525
01:29:25.373 --> 01:29:26.253
Yeah.

1526
01:29:26.414 --> 01:29:30.595
But like the measure that we have, which is this full width I have max, is like it's going to be the same for both.

1527
01:29:30.995 --> 01:29:31.335
Got it.

1528
01:29:31.994 --> 01:29:32.794
Right?

1529
01:29:32.934 --> 01:29:33.714
So which one is it?

1530
01:29:34.035 --> 01:29:37.775
And that's why there was a necessity for an independent line of evidence.

1531
01:29:38.476 --> 01:29:40.996
And this is where we're getting into the two papers that we have now.

1532
01:29:41.196 --> 01:29:41.836
Okay.

1533
01:29:42.176 --> 01:29:49.138
So it's this challenge of, well, it could be a defect, or it could be a truly transient object.

1534
01:29:49.518 --> 01:29:50.138
Which one is it?

1535
01:29:50.559 --> 01:30:00.441
And the interesting note here, just as a caveat or a point of order, is there was an emphasis on this being, you know,

1536
01:30:01.525 --> 01:30:05.548
The Polymar Sky Survey, which happened in the 50s, was pre-Sputnik.

1537
01:30:06.649 --> 01:30:07.689
Yes.

1538
01:30:07.809 --> 01:30:08.450
Sky Survey.

1539
01:30:09.551 --> 01:30:15.855
And nowadays, when we talk about transient light artifacts, it's...

1540
01:30:17.889 --> 01:30:39.811
starlink it's other satellites other whatever you chinese classified platforms but there's a lot of stuff yeah that we have now put up there up there and so it is much easier nowadays to say oh it's x y or z yep one of the interesting aspects of why this

1541
01:30:41.212 --> 01:30:51.000
distinguishing between whether it's a plate artifact or not, is if it's not, we're seeing these transient light anomalies that,

1542
01:30:52.572 --> 01:30:57.274
At a time period where we... We at least didn't put anything up there, right?

1543
01:30:57.594 --> 01:30:58.094
On purpose.

1544
01:30:58.154 --> 01:30:58.794
On purpose.

1545
01:30:58.994 --> 01:30:59.214
Right?

1546
01:30:59.714 --> 01:31:02.855
This is from November 1949 to December 1958.

1547
01:31:03.336 --> 01:31:04.896
This is before Sputnik.

1548
01:31:05.076 --> 01:31:06.197
Which could be interesting.

1549
01:31:06.357 --> 01:31:06.617
Yes.

1550
01:31:06.797 --> 01:31:12.659
Because of the fact that it can't just be pointed to being Starlink passing by.

1551
01:31:12.899 --> 01:31:13.379
Exactly.

1552
01:31:13.479 --> 01:31:13.919
Exactly.

1553
01:31:14.199 --> 01:31:18.941
And so now that is where we're getting into the two papers that have come out from the Vasco project, okay?

1554
01:31:18.961 --> 01:31:19.181
Okay.

1555
01:31:19.941 --> 01:31:35.937
What they've done is with the first paper, they're doing a spatial arrangement on individual plates, and they're arguing about the transients, how they look on individual plates, and where they are in the night sky in terms of...

1556
01:31:37.692 --> 01:31:53.167
when these photographic plates were taken, where was the Earth's shadow, and if we can do some kind of statistical comparison and really show that these are stuff that's in the night sky and not photographic plate defects.

1557
01:31:53.587 --> 01:31:59.473
And then there's a second paper that talks about temporal correlations in time between the...

1558
01:32:00.686 --> 01:32:07.496
Stuff that we see in the photographic plates and stuff like nuclear tests and UAP sightings, which we'll get to.

1559
01:32:07.996 --> 01:32:09.058
So, study one.

1560
01:32:09.458 --> 01:32:11.261
This is the alignment hypothesis.

1561
01:32:11.581 --> 01:32:14.705
This is Villa Royale and others in 2025 in PASP.

1562
01:32:14.725 --> 01:32:14.906
Yes.

1563
01:32:14.926 --> 01:32:15.226
Okay.

1564
01:32:18.142 --> 01:32:22.343
The paper is aligned multiple transient events in the first Palomar sky survey.

1565
01:32:22.683 --> 01:32:31.385
The hypothesis is that these multiple point-like transients are aligned along a straight line on the same plate, which is very unlikely by chance.

1566
01:32:31.945 --> 01:32:33.665
And they could be real physical objects.

1567
01:32:33.765 --> 01:32:35.406
And that is the argument that they're making.

1568
01:32:35.646 --> 01:32:35.886
Okay.

1569
01:32:36.186 --> 01:32:43.068
So they've got a catalog of about 300,000 short duration transients from this Palomar survey that we just went into.

1570
01:32:43.428 --> 01:32:43.608
Yes.

1571
01:32:43.928 --> 01:32:44.028
Okay.

1572
01:32:44.048 --> 01:32:44.208
Okay.

1573
01:32:45.288 --> 01:32:51.297
And the first thing that they do is they ask how many of these things are aligned.

1574
01:32:52.338 --> 01:32:53.520
They're in a line.

1575
01:32:53.800 --> 01:32:54.081
Okay.

1576
01:32:54.521 --> 01:32:58.467
And here we've got a photo of plate five.

1577
01:32:58.607 --> 01:32:59.748
This is candidate five.

1578
01:33:00.349 --> 01:33:01.931
There's five different thingies.

1579
01:33:02.996 --> 01:33:05.319
that are all in a line, okay?

1580
01:33:05.720 --> 01:33:16.634
And they use this thing, a statistical framework from Edmonds and George, published in 1985, and it calculates how often you would expect these things to happen by chance.

1581
01:33:16.914 --> 01:33:17.575
Okay.

1582
01:33:17.595 --> 01:33:17.856
Okay?

1583
01:33:18.196 --> 01:33:19.418
And they find a sigma of 3.9

1584
01:33:21.227 --> 01:33:22.949
That this is not chance.

1585
01:33:23.809 --> 01:33:24.069
Okay?

1586
01:33:24.090 --> 01:33:24.870
3.9 is...

1587
01:33:25.451 --> 01:33:26.051
It's pretty high.

1588
01:33:26.592 --> 01:33:27.172
Okay?

1589
01:33:27.292 --> 01:33:28.774
It's not like 5 Sigma.

1590
01:33:28.834 --> 01:33:39.202
Which we've talked about in our previous episode about the strongest evidence yet for life on an exoplanet, which was 3 Sigma.

1591
01:33:39.222 --> 01:33:40.203
Which was 3 Sigma.

1592
01:33:40.223 --> 01:33:40.804
It was 3.9.

1593
01:33:40.844 --> 01:33:42.745
It's higher, but it's still not 5 Sigma.

1594
01:33:42.785 --> 01:33:42.986
Sure.

1595
01:33:43.146 --> 01:33:43.326
Right?

1596
01:33:43.606 --> 01:33:44.087
But it's...

1597
01:33:45.622 --> 01:33:46.403
Kind of interesting.

1598
01:33:46.724 --> 01:33:46.964
Okay?

1599
01:33:47.224 --> 01:33:48.406
And this is just candidate five.

1600
01:33:48.446 --> 01:33:51.751
There's other candidates that also have around three sigma, right?

1601
01:33:52.132 --> 01:33:55.236
And they go through this in their paper.

1602
01:33:56.558 --> 01:33:59.623
That's actually not personally what I found the most interesting.

1603
01:33:59.663 --> 01:34:00.584
The alignment is like...

1604
01:34:01.164 --> 01:34:02.705
Like, I don't know, by chance.

1605
01:34:02.725 --> 01:34:08.328
I don't really know how to reason about that in my head, about how often to do things by chance.

1606
01:34:08.728 --> 01:34:22.495
And also the framework, the statistical framework by Edmonds and George, it relies on a bunch of different things like the number of objects in the sky, the assumption about the density of objects, how high they are, like all this other kind of stuff, right?

1607
01:34:22.515 --> 01:34:23.035
Right.

1608
01:34:23.215 --> 01:34:27.539
The Earth's shadow test is, I think, a decisive diagnostic.

1609
01:34:27.559 --> 01:34:28.379
So here's the logic.

1610
01:34:28.900 --> 01:34:34.004
You've got the Sun, you've got the Earth, and the Earth casts a shadow on the rest of space, right?

1611
01:34:34.344 --> 01:34:37.407
When the Moon goes in the Earth's shadow, that's how you get a lunar eclipse.

1612
01:34:38.828 --> 01:34:39.068
So...

1613
01:34:41.589 --> 01:34:54.002
If we were to think that all of these things are plate defects, all of these transients are plate defects and not actual objects in the sky that are reflecting sunlight, then...

1614
01:34:55.005 --> 01:35:03.007
If there are plate defects, the number of things that are happening in Earth's shadow should be the same as outside Earth's shadow.

1615
01:35:03.267 --> 01:35:03.407
Right.

1616
01:35:03.427 --> 01:35:06.027
Because it's not dependent on whether the thingy...

1617
01:35:06.067 --> 01:35:06.247
Right.

1618
01:35:06.287 --> 01:35:10.388
Whether the patch of sky I'm looking at is in Earth's shadow or not.

1619
01:35:10.568 --> 01:35:10.768
Right.

1620
01:35:11.088 --> 01:35:20.050
If, on the other hand, these objects are reflecting the light from the sun, then there should be fewer transients...

1621
01:35:20.868 --> 01:35:25.072
When I'm looking in Earth's shadow than when I'm not.

1622
01:35:25.353 --> 01:35:25.533
Right.

1623
01:35:25.873 --> 01:35:26.153
Okay?

1624
01:35:26.654 --> 01:35:34.002
And I know the coordinate of the spot in the sky where I'm looking, right?

1625
01:35:34.102 --> 01:35:40.488
Because each of these photographic plates has a right ascension and declination, which is our latitude and longitude in the sky.

1626
01:35:40.889 --> 01:35:41.950
We have our time of day.

1627
01:35:43.022 --> 01:35:49.286
So from that, I can calculate which patch of the photographic plate is in Earth's shadow and which patch is not.

1628
01:35:49.586 --> 01:35:53.729
I do have to do an assumption based on how high I think these objects are.

1629
01:35:54.029 --> 01:36:04.977
But they make an argument, which I actually don't quite fully understand, but they make an argument that these objects are probably at geosynchronous orbit, which is 42,000 kilometers.

1630
01:36:07.738 --> 01:36:08.559
And at that point...

1631
01:36:10.045 --> 01:36:17.391
I can calculate which part of my photographic plate is going to be in its shadow and which part is not.

1632
01:36:18.292 --> 01:36:25.758
And then from that, I get an expected number of 1,200 objects, but I only observe 349.

1633
01:36:27.279 --> 01:36:31.302
So that's a pretty statistically significant sample.

1634
01:36:31.422 --> 01:36:38.207
And this is where I remember when you first came to me with this story, you said 22 sigma significance, right?

1635
01:36:39.248 --> 01:37:01.900
the statistical significance i was like what are you talking about this is what they're talking about right they're talking about that like the 349 is so far away from 1223 it's 22 sigma away they actually did a more realistic comparison where what they did was um they they looked at the actual sky because this was actually kind of cool so the the photographic plates

1636
01:37:03.005 --> 01:37:08.269
in Palomar didn't uniformly sample the entire night sky.

1637
01:37:08.309 --> 01:37:14.753
The 22 Sigma comes from the assumption that the photographic plates are completely uniformly sampling the night sky.

1638
01:37:15.113 --> 01:37:18.696
On the other hand, some patches were looked at more than others, right?

1639
01:37:18.736 --> 01:37:20.397
And some patches were completely missed.

1640
01:37:20.737 --> 01:37:28.443
So what they did was they did a more realistic simulation where they said, okay, suppose I only looked at those patches and I randomly generated objects.

1641
01:37:29.489 --> 01:37:33.914
How often would I see defects versus not defects and all that other kind of stuff?

1642
01:37:34.294 --> 01:37:36.717
And there they get a lower significance, but it's still a lot.

1643
01:37:36.757 --> 01:37:38.359
7.6 sigma.

1644
01:37:39.279 --> 01:37:40.421
That sits better with me.

1645
01:37:41.382 --> 01:37:41.602
Okay?

1646
01:37:41.622 --> 01:37:41.862
7.6?

1647
01:37:41.882 --> 01:37:42.042
Yeah.

1648
01:37:42.082 --> 01:37:42.343
Seems...

1649
01:37:45.051 --> 01:37:45.691
more reasonable.

1650
01:37:46.331 --> 01:37:49.353
And, and the benchmarking that they did kind of made more sense to me, right?

1651
01:37:49.373 --> 01:37:58.416
Because they're, they're taking into account this prior distribution of like where the telescope was looking rather than an assumption that, Oh, it just looked everywhere.

1652
01:37:58.696 --> 01:37:59.216
Well, it didn't.

1653
01:37:59.536 --> 01:37:59.736
Right.

1654
01:37:59.796 --> 01:38:01.657
It's not the Vera Rubin observatory.

1655
01:38:01.717 --> 01:38:01.937
Right.

1656
01:38:01.977 --> 01:38:02.177
Right.

1657
01:38:02.197 --> 01:38:03.557
It's not just like, yeah, yeah.

1658
01:38:03.597 --> 01:38:04.478
This is the fifties.

1659
01:38:04.518 --> 01:38:06.318
Yeah.

1660
01:38:06.699 --> 01:38:09.860
It took us 80 years to get, to get to the Vera Rubin.

1661
01:38:09.920 --> 01:38:10.180
So, yeah,

1662
01:38:11.540 --> 01:38:13.601
So that's the first story.

1663
01:38:13.641 --> 01:38:21.746
The second story, which is study number two, this is published in Nature Scientific Reports, and this is correlations with the atomic age.

1664
01:38:23.624 --> 01:38:32.811
Transients in the Palomar Observatory Sky Survey may be associated with nuclear testing and reports of unidentified anomalous phenomenon.

1665
01:38:33.251 --> 01:38:35.253
This is, it's got UAP in the title.

1666
01:38:35.293 --> 01:38:36.334
It's got UAP in the title.

1667
01:38:36.414 --> 01:38:37.274
Bro.

1668
01:38:37.474 --> 01:38:38.495
In a nature subjournal.

1669
01:38:38.575 --> 01:38:39.556
In a nature subjournal.

1670
01:38:39.576 --> 01:38:41.037
Springer Publication.

1671
01:38:41.077 --> 01:38:48.142
Which is a significant thing in and of itself given the aversion of...

1672
01:38:48.943 --> 01:38:53.085
to doing anything related to this subject for most peer-reviewed journals.

1673
01:38:53.305 --> 01:38:55.266
Yeah, I mean, I was pretty surprised myself.

1674
01:38:55.286 --> 01:38:57.767
Yeah, when you were like, because I brought it up, you're like, well, what journal is it?

1675
01:38:57.787 --> 01:38:59.147
And I was like, oh, and you're like, wait, really?

1676
01:38:59.167 --> 01:39:00.388
Wait, nature, yeah.

1677
01:39:00.588 --> 01:39:03.729
Fine, I guess if it's Springer, I thought we could do it.

1678
01:39:04.149 --> 01:39:05.330
We'll cover it if it's Springer.

1679
01:39:05.370 --> 01:39:06.130
Yeah, we'll cover it.

1680
01:39:06.170 --> 01:39:10.032
So here what they're doing is they're doing more of a time-

1681
01:39:10.978 --> 01:39:12.059
Sensitive approach, okay?

1682
01:39:12.859 --> 01:39:18.482
They're looking at the rate of the transients that appear in those photographic plates that I talked about.

1683
01:39:18.862 --> 01:39:27.567
And they're trying to correlate it with dates of above-ground nuclear weapons tests and daily counts of UAP reports, okay?

1684
01:39:28.148 --> 01:39:30.449
So the timeline is, again, 1949 to 1957.

1685
01:39:30.489 --> 01:39:30.649
Okay.

1686
01:39:33.149 --> 01:39:35.930
They get a bunch of transient count.

1687
01:39:36.210 --> 01:39:42.292
So for every single day, they've got photographic plates, and they count how many transients were found on each day.

1688
01:39:43.572 --> 01:39:49.714
Each day also gets a binary flag on whether it's within a day of a nuclear test or not.

1689
01:39:49.914 --> 01:39:50.114
Right?

1690
01:39:50.434 --> 01:39:55.696
And back then, I think only three countries were doing nuclear tests, the Soviets, the U.S., and Great Britain.

1691
01:39:56.336 --> 01:40:01.358
And then there's also daily UAP reports from the UFO CAT database.

1692
01:40:02.723 --> 01:40:06.265
And these three things are what they're trying to relate.

1693
01:40:07.286 --> 01:40:09.727
They use something called the generalized linear model.

1694
01:40:09.827 --> 01:40:14.730
I'm very familiar with something called the generalized linear model because my PhD had a lot of that.

1695
01:40:15.171 --> 01:40:23.736
Basically, these kinds of models are what you want to do when things are not just simply linearly correlated to one another.

1696
01:40:24.776 --> 01:40:27.078
In my PhD specifically, I was trying to...

1697
01:40:28.847 --> 01:40:37.429
correlate the spiking of neurons in terms of many different behavioral motifs, like which way he's looking, how fast he's running, where he is.

1698
01:40:37.989 --> 01:40:43.890
All of these things can have varied responses that are not necessarily like, oh, if I'm looking this way, I just have more.

1699
01:40:44.210 --> 01:40:46.131
Maybe it's like a weird response, right?

1700
01:40:46.171 --> 01:40:48.811
Where only if he's looking this way, there's going to be more.

1701
01:40:49.191 --> 01:40:55.293
So all of these different multimodal effects can be captured by a generalized linear model.

1702
01:40:55.333 --> 01:40:56.073
And that's what they did.

1703
01:40:57.561 --> 01:41:03.651
It can also simulate weird distributions that aren't just uniformly distributed or Gaussian distributed.

1704
01:41:03.691 --> 01:41:07.517
If you've got a binomial distribution, which is like coin flips.

1705
01:41:07.777 --> 01:41:08.599
Yep.

1706
01:41:08.719 --> 01:41:11.303
If you've got, you know, did I see something or did I not?

1707
01:41:11.623 --> 01:41:12.204
Things like that.

1708
01:41:13.097 --> 01:41:15.399
It's very good at modeling those kinds of things, right?

1709
01:41:15.939 --> 01:41:18.461
So there was a correlation to nuclear testing.

1710
01:41:19.102 --> 01:41:26.467
So the transience occurrence in these photographic plates was 45% more likely within a three-day nuclear test window.

1711
01:41:27.107 --> 01:41:27.348
Okay?

1712
01:41:27.848 --> 01:41:32.612
And the main thing was it was 68% higher likelihood on a day after the test.

1713
01:41:33.693 --> 01:41:33.913
Right?

1714
01:41:34.533 --> 01:41:35.434
It's pretty significant.

1715
01:41:35.454 --> 01:41:36.295
It's pretty significant.

1716
01:41:36.535 --> 01:41:49.208
There's also a correlation plot that they do with UAP reports where using that generalized linear model analysis, 8.5% increase in transients for every additional independent UAP report on the same day.

1717
01:41:49.669 --> 01:41:51.490
I don't really care for that as much.

1718
01:41:51.711 --> 01:41:53.713
Like UAP reports to me are like...

1719
01:41:54.660 --> 01:41:54.940
I don't know.

1720
01:41:54.980 --> 01:41:56.782
Humans are fooled very easily.

1721
01:41:56.962 --> 01:42:00.686
And UAP reports are just humans being like, I saw something that was crazy.

1722
01:42:02.548 --> 01:42:09.214
I can't think of whether there are tests that I could do to see if this would work.

1723
01:42:09.234 --> 01:42:15.340
For example, what if you assumed that on average, half of the UAP reports were nonsense?

1724
01:42:16.040 --> 01:42:17.282
If you took out half,

1725
01:42:17.961 --> 01:42:20.043
would the correlation still be there?

1726
01:42:20.423 --> 01:42:29.009
If you did a bunch of samples where you took out half and you looked for correlation, would there be some samples where the correlation would be way higher?

1727
01:42:29.470 --> 01:42:30.651
Yeah.

1728
01:42:30.691 --> 01:42:34.113
That's something that we used to do in my neuroscience lab in PhD.

1729
01:42:34.153 --> 01:42:34.454
It's like...

1730
01:42:35.014 --> 01:42:39.097
How sure are we that this variable is correlated to this variable?

1731
01:42:39.358 --> 01:42:47.404
Where if I were to just take out a subset of the data, if the correlation still persists, then it's a more robust finding, right?

1732
01:42:47.725 --> 01:42:48.265
Yep.

1733
01:42:48.285 --> 01:42:51.307
And given the shape of that plot, I'm not sure if I believe it.

1734
01:42:51.808 --> 01:42:57.673
But the nuclear testing thing is just like, that's just there, right?

1735
01:42:57.713 --> 01:42:58.053
Yeah, yeah.

1736
01:42:59.414 --> 01:43:01.656
So those are the two reports.

1737
01:43:02.077 --> 01:43:04.099
There's one report in PASP.

1738
01:43:04.699 --> 01:43:05.860
That's physical evidence.

1739
01:43:06.221 --> 01:43:10.545
You've got the Earth's shadow test that a substantial fraction of the transients are...

1740
01:43:12.030 --> 01:43:13.751
sunlit objects, right?

1741
01:43:13.831 --> 01:43:25.059
And then you've got the scientific reports, that's circumstantial behavioral evidence that says that the timing of these transients is affected by nuclear tests, right?

1742
01:43:26.660 --> 01:43:29.022
There are people who obviously...

1743
01:43:31.765 --> 01:43:33.146
are skeptics, right?

1744
01:43:33.306 --> 01:43:46.335
As we also saw with the same paper I just brought up earlier about the strongest evidence yet, that came immediately with follow-ups from multiple institutions that pointed to why their modeling was not sufficiently robust.

1745
01:43:46.355 --> 01:43:46.615
Yeah.

1746
01:43:47.016 --> 01:43:51.078
Like, as of now, I haven't seen any follow-up scientific papers.

1747
01:43:51.579 --> 01:43:51.919
Right.

1748
01:43:52.159 --> 01:43:56.142
Because all of this data is publicly available.

1749
01:43:56.202 --> 01:43:57.583
And I actually saw on the...

1750
01:43:58.846 --> 01:44:02.968
On the two papers, it's readily available for people to look at.

1751
01:44:03.008 --> 01:44:06.730
So I encourage anyone to do their own analysis, actually.

1752
01:44:07.430 --> 01:44:08.431
I want to make a funny note.

1753
01:44:09.011 --> 01:44:12.733
On the paper that I keep bringing up, within two weeks, because it was, was it?

1754
01:44:13.413 --> 01:44:14.154
It was a week, dude.

1755
01:44:14.194 --> 01:44:14.674
It was a week.

1756
01:44:14.734 --> 01:44:14.934
Yeah.

1757
01:44:14.954 --> 01:44:17.615
Within a week, because it was out of Cambridge or Oxford, one of the two.

1758
01:44:17.655 --> 01:44:20.817
Yeah, Cambridge had the original paper, and then the guy at Oxford was like, nah.

1759
01:44:20.837 --> 01:44:22.938
And within a week, Oxford came out with the rebuttal.

1760
01:44:22.978 --> 01:44:23.198
Yeah.

1761
01:44:23.398 --> 01:44:24.519
And then multiple came out.

1762
01:44:24.959 --> 01:44:26.220
It's been more than a week.

1763
01:44:26.300 --> 01:44:26.400
Yeah.

1764
01:44:26.420 --> 01:44:26.600
Yeah.

1765
01:44:27.993 --> 01:44:35.576
This paper is getting quite a bit of just, obviously, a lot of the science media is doing a variety of different coverage.

1766
01:44:35.837 --> 01:44:38.398
But there's not yet necessarily been follow-ups as quickly.

1767
01:44:39.198 --> 01:44:40.699
There's a number of reasons that could be true.

1768
01:44:40.939 --> 01:44:46.441
Yeah, I mean, one of them, I think the big one is probably it's not an active scientific field.

1769
01:44:46.541 --> 01:44:52.304
Like exoplanets and exoplanet researches, everyone's trying to make a name for themselves, right?

1770
01:44:52.324 --> 01:44:53.644
They've got a career to look after.

1771
01:44:54.044 --> 01:44:54.925
This one is not like...

1772
01:44:55.600 --> 01:44:58.201
Something that, you know, okay, so you proved this wrong.

1773
01:44:58.261 --> 01:45:01.322
It's not going to, like, give you a leg up in an astrophysical career.

1774
01:45:01.443 --> 01:45:05.924
There's no incentive for people who are looking to climb a ladder to do a follow-up on it.

1775
01:45:05.944 --> 01:45:07.145
Yeah, yeah, yeah.

1776
01:45:07.285 --> 01:45:08.546
So that could be one of the reasons.

1777
01:45:10.815 --> 01:45:19.200
But in any case, the Scientific American actually had an article about it that I was reading, and they showed some rebuttals, right?

1778
01:45:19.681 --> 01:45:23.223
So there's the Hambly argument, which is from the University of Edinburgh.

1779
01:45:23.903 --> 01:45:26.565
He's an expert, Nigel Hambly.

1780
01:45:26.585 --> 01:45:28.706
He's an expert in photographic plate analysis.

1781
01:45:29.467 --> 01:45:29.647
And...

1782
01:45:30.327 --> 01:45:32.888
His main contention was that full width at half max, right?

1783
01:45:32.948 --> 01:45:41.173
The fact that these transients are so pinpointy and precise compared to the stars that are kind of spread out because of all the noise from the detectors.

1784
01:45:41.733 --> 01:45:50.018
And there's skepticism that, you know, this is a classic signature of an emulsion or a flaw or a plate defect.

1785
01:45:50.338 --> 01:45:54.200
And it's not a celestial object because a celestial object would have this kind of...

1786
01:45:54.882 --> 01:45:56.043
noise.

1787
01:45:56.423 --> 01:46:01.486
And one of the crucial things that he says is the ultimate arbiter is actually just the physical evidence.

1788
01:46:01.927 --> 01:46:03.208
The plates are still around.

1789
01:46:03.708 --> 01:46:10.152
We can take those original glass plates and examine them under a microscope and not examine the digital scans.

1790
01:46:10.512 --> 01:46:21.480
And under a microscope, it's going to be very obvious whether this is a defect of the silver just being kind of weird at that point, or if it's a genuine astrophysical source.

1791
01:46:22.486 --> 01:46:27.610
So I think that's, you know, he's proposing a way to... To make it go away.

1792
01:46:27.650 --> 01:46:28.471
To make it go away.

1793
01:46:29.091 --> 01:46:29.792
I think that's good.

1794
01:46:30.272 --> 01:46:39.440
So if you're interested in doing a microscopic study of the digital plates and you would like to collaborate, just in the comments below.

1795
01:46:39.760 --> 01:46:40.861
Yeah, let us know.

1796
01:46:40.941 --> 01:46:41.482
Let us know.

1797
01:46:42.022 --> 01:46:47.406
And we already talked just previously about, even in the paper itself...

1798
01:46:48.463 --> 01:46:52.986
a means by which that is not necessarily the likely outcome.

1799
01:46:53.286 --> 01:46:55.167
However, you can remove it from the table.

1800
01:46:55.187 --> 01:46:55.767
The earth shadow thing.

1801
01:46:55.907 --> 01:46:56.988
Yes, the earth shadow point.

1802
01:46:57.488 --> 01:47:02.071
You can still remove it entirely from being the cudgel that you're seeing being put out.

1803
01:47:02.491 --> 01:47:12.837
Because a lot of times in the coverage of it, people say, oh, it's probably plate defects, without actually explaining how the paper points to why it's not plate defects.

1804
01:47:12.857 --> 01:47:13.517
Yeah, yeah, yeah.

1805
01:47:13.537 --> 01:47:14.718
Which is kind of the issue.

1806
01:47:14.738 --> 01:47:16.199
The earth shadow thing is pretty...

1807
01:47:17.357 --> 01:47:18.417
Pretty substantive, I think.

1808
01:47:18.777 --> 01:47:27.179
And so, however, the beauty of us still having the original plates is that it's falsifiable.

1809
01:47:27.319 --> 01:47:28.239
Yes, exactly.

1810
01:47:29.399 --> 01:47:33.480
There's another one that's proposed by Sean Kirkpatrick.

1811
01:47:33.920 --> 01:47:36.721
Ooh, Dr. Sean Kirkpatrick.

1812
01:47:37.141 --> 01:47:39.522
Former Pentagon AARO head.

1813
01:47:40.122 --> 01:47:41.062
We like to call it ARO.

1814
01:47:41.082 --> 01:47:41.242
Yeah.

1815
01:47:41.742 --> 01:47:42.063
Arrow.

1816
01:47:42.383 --> 01:47:43.264
Okay, Arrowhead.

1817
01:47:45.346 --> 01:47:47.769
I guess that's like an anomaly office or something?

1818
01:47:48.150 --> 01:47:50.212
That's the all-domain anomaly resolution office.

1819
01:47:50.232 --> 01:47:50.933
Oh, okay, got it.

1820
01:47:50.973 --> 01:47:57.861
That was created in 2022 as a result of Congress pushing on the executive branch to give them more information about... Again, because...

1821
01:47:59.142 --> 01:48:08.352
military and intelligence officers were coming to Congress saying, this is happening and we need help in dealing with this and we're not getting the support we need.

1822
01:48:08.493 --> 01:48:09.193
I see.

1823
01:48:09.253 --> 01:48:12.337
That was the impetus for why the office was established.

1824
01:48:12.917 --> 01:48:16.481
And its mandate was twofold, which was unfortunate because –

1825
01:48:17.302 --> 01:48:21.325
it was basically a mandate that had conflating outcomes.

1826
01:48:21.965 --> 01:48:34.593
One was we want you, Arrow, to take data from the Department of Defense, all of the combatant commands, and the intelligence community, intake it, and then analyze what we're seeing as these UAP reports.

1827
01:48:34.633 --> 01:48:37.915
And if it's conventional, send it to the office that exists to deal with it.

1828
01:48:38.236 --> 01:48:42.518
And if it's unconventional, figure out what we need to do to figure out what it is.

1829
01:48:43.299 --> 01:48:44.360
That was one of its mandates.

1830
01:48:44.380 --> 01:48:44.620
Yeah.

1831
01:48:45.140 --> 01:48:58.174
The other mandate was a public communications mandate to help to demystify this issue to the public by creating public education materials and this and the third.

1832
01:48:58.454 --> 01:49:05.662
The problem with that is you're asking the Department of Defense, which is not a public communications office, to do public communications.

1833
01:49:05.922 --> 01:49:07.644
There's a reason why NASA – Yeah.

1834
01:49:08.184 --> 01:49:09.305
Is its own agency.

1835
01:49:09.826 --> 01:49:15.891
And you don't have NASA running out of the DOD because we would never get anything out of that structure.

1836
01:49:15.951 --> 01:49:17.413
So there's sort of a conflate.

1837
01:49:17.793 --> 01:49:29.484
There's a friction between the two things Congress asked it to do, which is why a lot of the public is frustrated because they are not very forthcoming with – they have thousands of these reports.

1838
01:49:29.564 --> 01:49:30.605
We've basically seen none of it.

1839
01:49:30.885 --> 01:49:31.826
Yeah, okay.

1840
01:49:32.246 --> 01:49:45.038
Well, he put his, Sean Kirkpatrick, he gave his two cents, which I thought was actually quite good, which was, he said, you know, there's nuclear detonation, and clearly this thing is correlated with nuclear detonation.

1841
01:49:45.558 --> 01:49:51.303
Now, when you have these high-altitude nuclear detonations, you know, you can have radiation, particulate.

1842
01:49:52.244 --> 01:50:08.027
stuff that goes away and it can create transient luminous events like churn-carve radiation or even reflective metallic junk, like the stuff that made the nuclear, like, you know, there's going to be stuff that gets thrown out and that junk could be,

1843
01:50:08.748 --> 01:50:11.730
reflect light in the way that we're seeing in these Palomar sky surveys.

1844
01:50:12.350 --> 01:50:23.077
This is interesting because I was actually, I was just at the Seoul conference in Lake Majore, Italy this past week, where Dr. Bischoff-Viller-Real presented these two papers to,

1845
01:50:24.437 --> 01:50:33.239
And one of the – there was a Q&amp;A section where one of the people actually asked about the Chernkov radiation as a solution to the problem set.

1846
01:50:33.679 --> 01:50:36.440
Someone can clip this and put the response.

1847
01:50:36.460 --> 01:50:37.520
I don't think it's public yet.

1848
01:50:37.940 --> 01:50:42.442
I didn't write it down, but she did have a response to that issue.

1849
01:50:42.662 --> 01:50:43.402
Yeah.

1850
01:50:43.462 --> 01:50:46.182
Which I unfortunately can't replicate in this context.

1851
01:50:46.583 --> 01:50:52.324
But that was brought up by someone in the audience as the obvious next step in the progression.

1852
01:50:52.344 --> 01:50:52.444
Yeah.

1853
01:50:52.564 --> 01:50:53.044
Yeah.

1854
01:50:53.124 --> 01:50:53.905
As an answer to it.

1855
01:50:54.385 --> 01:51:01.570
Yeah, I can imagine, like, I mean, for Cherkhov radiation, it would probably be, like, streaky rather than single point sources.

1856
01:51:03.132 --> 01:51:11.818
But, you know, if you have, like, metallic junk, then that could, like, just, like, you know, as it's turning, like, create that reflective.

1857
01:51:12.178 --> 01:51:21.385
So, to me, the metallic junk is a better explanation than Cherkhov radiation or ionization, because those have very characteristic...

1858
01:51:22.203 --> 01:51:24.464
signatures on photographic plates.

1859
01:51:24.904 --> 01:51:29.966
But one of the challenges that Sean Kirkpatrick said was, you know, we've got GPS satellites now.

1860
01:51:30.386 --> 01:51:35.688
Why don't we try to reproduce these results with the GPS satellites?

1861
01:51:37.249 --> 01:51:38.169
I think it would...

1862
01:51:39.713 --> 01:51:43.474
I think it's harder said than done because GPS, there's so many satellites now.

1863
01:51:43.494 --> 01:51:44.135
Yeah, yeah, yeah.

1864
01:51:44.195 --> 01:51:44.375
Right?

1865
01:51:44.415 --> 01:51:46.735
It's not just like you're going to have so many streaks.

1866
01:51:46.775 --> 01:51:47.456
Yeah, yeah.

1867
01:51:48.336 --> 01:51:52.958
So for a 50-minute exposure, you're just going to get completely wrecked by Starlink.

1868
01:51:53.178 --> 01:51:53.618
Right.

1869
01:51:53.778 --> 01:52:02.481
There's another proposition by Elliot Gillum at the SETI Institute, Search for Extraterrestrial Intelligence.

1870
01:52:02.521 --> 01:52:05.502
He said, you know, meteors traveling directly towards the telescope

1871
01:52:06.182 --> 01:52:08.864
Could appear stationary dots, not streaks.

1872
01:52:09.544 --> 01:52:14.667
Um, it's viable, but like now you got, it's traveling directly at the telescope.

1873
01:52:14.727 --> 01:52:17.529
Like that, that, that volume level and simultaneous.

1874
01:52:17.549 --> 01:52:17.749
Yeah.

1875
01:52:17.769 --> 01:52:17.969
Yeah.

1876
01:52:18.010 --> 01:52:18.150
Yeah.

1877
01:52:18.210 --> 01:52:20.811
It's like simultaneous, multiple aligned.

1878
01:52:20.831 --> 01:52:22.272
It's a little challenging.

1879
01:52:22.292 --> 01:52:22.392
Yeah.

1880
01:52:22.412 --> 01:52:24.273
It's a little challenging given the numbers.

1881
01:52:24.854 --> 01:52:25.034
Um, but,

1882
01:52:26.358 --> 01:52:30.440
There's a Princeton astrophysicist, Robert Lupton, who has this look elsewhere effect.

1883
01:52:30.820 --> 01:52:41.305
He's basically saying that, you know, in a massive data set with thousands of plates, millions of images, finding a few apparent alignments by chance is not that surprising.

1884
01:52:41.605 --> 01:52:48.409
And you can't actually use the Edmonds and George formula to do that kind of statistical significance.

1885
01:52:49.189 --> 01:52:53.192
to figure out whether stuff isn't aligned because there's so many things that are in a line.

1886
01:52:53.752 --> 01:53:01.057
You can easily fool yourself into thinking that the thing that isn't aligned is there for some reason, right?

1887
01:53:01.117 --> 01:53:06.040
And the probabilities need to be corrected for the vast number of stuff that's out there.

1888
01:53:06.340 --> 01:53:09.802
This goes back to your not looking at the alignment as the interesting part.

1889
01:53:09.842 --> 01:53:10.843
Yeah, yeah, yeah.

1890
01:53:10.943 --> 01:53:17.090
And I think there's a, I need a little bit more convincing on the alignment being something that's interesting.

1891
01:53:17.250 --> 01:53:17.530
Okay.

1892
01:53:17.610 --> 01:53:18.672
Right.

1893
01:53:19.152 --> 01:53:26.960
But I mean, the debate is there, you know, 7.6 sigma deficit of transience in the Umbra.

1894
01:53:27.341 --> 01:53:29.563
That's not something that I think we can ignore.

1895
01:53:31.305 --> 01:53:41.496
I think it forces a shift from are any of these transients real to, okay, given a significant fraction of them are probably real sunlit objects, what are they?

1896
01:53:42.097 --> 01:53:42.357
Okay.

1897
01:53:44.439 --> 01:53:48.643
It's certainly like giving me pause.

1898
01:53:48.664 --> 01:53:48.924
Yeah.

1899
01:53:49.184 --> 01:53:49.744
I'll be honest.

1900
01:53:49.884 --> 01:53:54.106
Which I want to say for the audience, I've been working this guy for a long time.

1901
01:53:54.686 --> 01:54:00.248
So to be giving him pause is statistically significant.

1902
01:54:00.268 --> 01:54:02.870
Yeah, yeah, yeah, yeah, yeah.

1903
01:54:02.890 --> 01:54:06.751
You probably have, like the number of times you've tried, I'm like, nah.

1904
01:54:06.891 --> 01:54:08.112
I've tried so hard.

1905
01:54:08.292 --> 01:54:15.635
But I thought you'd find this interesting, even just from the experimental design perspective, the clever approach of trying to do a study like this

1906
01:54:16.635 --> 01:54:17.737
In a way that's compelling.

1907
01:54:17.978 --> 01:54:19.420
Yeah.

1908
01:54:19.701 --> 01:54:22.246
I mean, clearly, like, the team...

1909
01:54:23.727 --> 01:54:37.850
the team in the PASP paper, at least the, the one in the proceedings of the Pacific astronomical society, that one, that one has done its job of like, you know, going through the proper science and things like that.

1910
01:54:38.170 --> 01:54:49.692
The, the time stuff I'm a little less excited about, but the, the, the one with the, the earth shadow thing, that's just like solid calculation, you know, good math.

1911
01:54:49.772 --> 01:54:49.973
Yeah.

1912
01:54:49.993 --> 01:54:50.953
That's just like good math.

1913
01:54:51.433 --> 01:54:51.573
Um,

1914
01:54:52.873 --> 01:54:55.174
There is one that was kind of interesting.

1915
01:54:55.194 --> 01:55:01.216
So Candidate 5, the five transients systematically in alignment, that happened on July 22nd, 1952.

1916
01:55:01.236 --> 01:55:01.376
27th.

1917
01:55:01.576 --> 01:55:01.897
Oh, yeah.

1918
01:55:01.917 --> 01:55:03.477
July 27th, 1952.

1919
01:55:03.497 --> 01:55:04.918
And they mentioned that this coincides with the...

1920
01:55:12.120 --> 01:55:27.625
most intense weekend of 1952 Washington, D.C. UFO flap, which I hadn't heard about, but I looked it up, and we've got a photo showing the recorded number of incidents that were reported.

1921
01:55:27.905 --> 01:55:35.068
And apparently there were visual sightings, fighter jets were scrambled, there was a press conference from the Pentagon.

1922
01:55:35.908 --> 01:55:38.469
So I didn't even know that this thing happened.

1923
01:55:38.649 --> 01:55:40.310
So this is one of the weird, bro.

1924
01:55:40.350 --> 01:55:41.971
This is one of the things that's so interesting.

1925
01:55:41.991 --> 01:55:45.253
Why would they do that?

1926
01:55:46.493 --> 01:55:57.159
It's, it's so interesting because there's, again, this is why when we start at the beginning of this, there's this whole legacy of this subject that exists.

1927
01:55:57.419 --> 01:55:58.259
That's very dense.

1928
01:55:58.459 --> 01:55:59.960
And there's a lot of information there.

1929
01:56:00.700 --> 01:56:03.302
And, you know, unless you have an entry point,

1930
01:56:04.596 --> 01:56:15.724
You won't know it's there, but UFO flaps are consistent and persistent throughout the history, not only of the US, but global society.

1931
01:56:15.744 --> 01:56:21.388
And it is interesting to see how it cyclically comes back up.

1932
01:56:21.648 --> 01:56:22.789
But that was Candidate 5.

1933
01:56:23.850 --> 01:56:26.251
Yeah, and that was the one that was super aligned.

1934
01:56:26.711 --> 01:56:32.714
I mean, I still think it's probably coincidence, but it's an interesting coincidence.

1935
01:56:32.774 --> 01:56:34.435
I'll say that.

1936
01:56:34.535 --> 01:56:38.157
The other thing they do at the end of the PASP paper...

1937
01:56:40.658 --> 01:57:00.791
is they look at 3D simulations in Blender and show how these reflective objects, if they're like slowly processing something that has a shape like that, could create the kind of signature that we see where like randomly it's at exactly the right angle where it's reflecting the sun into the Palomar telescope and then it goes away real quick.

1938
01:57:00.811 --> 01:57:06.435
So then you get a really sharp image spec in your photographic plate.

1939
01:57:08.751 --> 01:57:12.972
It's speculative at best.

1940
01:57:13.112 --> 01:57:15.793
I think it's certainly interesting.

1941
01:57:15.873 --> 01:57:20.894
I think the Earth's shadow thing is the one that is really sticking with me.

1942
01:57:20.954 --> 01:57:22.554
That calculation is quite nice.

1943
01:57:22.854 --> 01:57:33.677
And the way that they do the controls, where they simulate a random subset in that photographic plate and say, okay, how many are going to be in Earth's shadow versus not?

1944
01:57:33.977 --> 01:57:34.437
Yeah, that's...

1945
01:57:35.497 --> 01:57:41.999
Seems to be the way that I used to do, like, you know, these bootstrap things to try to figure out what the significance of my result was.

1946
01:57:42.519 --> 01:57:44.760
So, yeah, I thought it was pretty cool.

1947
01:57:44.800 --> 01:57:47.901
I think next steps would be to just look at the photographic plate.

1948
01:57:48.541 --> 01:57:49.801
Make sure that it's not a defect.

1949
01:57:49.981 --> 01:57:52.102
Take a look at the plate, not a defect.

1950
01:57:52.142 --> 01:57:57.644
And then from there, if proven to be the case, now we can have the next stage of the conversation.

1951
01:57:57.664 --> 01:58:01.725
Yeah, which is like, you know, with modern surveys, perhaps there's a way to get...

1952
01:58:02.630 --> 01:58:04.952
An idea of how often this is happening in Verirubin.

1953
01:58:04.972 --> 01:58:09.855
Verirubin's completely public.

1954
01:58:09.935 --> 01:58:12.497
All of the data is public immediately after it's taken.

1955
01:58:12.517 --> 01:58:13.657
And it's up and running right now.

1956
01:58:13.677 --> 01:58:14.078
Yeah, yeah.

1957
01:58:14.098 --> 01:58:17.320
So we have already three months of data from the Verirubin.

1958
01:58:17.460 --> 01:58:23.684
If you've not already listened, take a look back to our previous episode where we covered the Verirubin Observatory.

1959
01:58:24.265 --> 01:58:25.145
Why it matters.

1960
01:58:25.225 --> 01:58:27.167
It's another sky survey, but it is...

1961
01:58:27.847 --> 01:58:36.429
Just light years in terms of the structure and substance of how it works and what we talked about, the immediate data availability.

1962
01:58:36.869 --> 01:58:37.669
No embargo.

1963
01:58:38.190 --> 01:58:44.611
Again, some people will be like, well, you know, government's going to try to wipe away their classified projects that are –

1964
01:58:48.492 --> 01:58:49.952
Let's say three months.

1965
01:58:50.253 --> 01:58:50.473
Yeah.

1966
01:58:50.613 --> 01:58:55.234
Three months meaning 100 days, meaning there's 30 photos of the entire night sky already.

1967
01:58:55.394 --> 01:58:55.574
Right.

1968
01:58:55.794 --> 01:58:56.134
Right.

1969
01:58:56.174 --> 01:58:59.675
Palomar took years for one photo of the entire night sky.

1970
01:58:59.695 --> 01:59:00.876
This already has 30.

1971
01:59:01.396 --> 01:59:04.197
It's an incredible, the largest...

1972
01:59:04.937 --> 01:59:05.217
Camera?

1973
01:59:05.237 --> 01:59:05.398
Yeah.

1974
01:59:05.778 --> 01:59:06.879
The largest camera in the world.

1975
01:59:06.919 --> 01:59:07.519
In the world.

1976
01:59:09.361 --> 01:59:11.642
Creating the biggest amount of data.

1977
01:59:11.783 --> 01:59:11.963
Yeah.

1978
01:59:12.023 --> 01:59:16.266
And one of the most incredible data processing and distribution platforms.

1979
01:59:16.486 --> 01:59:16.686
Yeah.

1980
01:59:16.987 --> 01:59:17.647
It's insane.

1981
01:59:17.787 --> 01:59:18.788
It's just incredible.

1982
01:59:19.148 --> 01:59:31.298
So the point here being there are things that can be drawn from this study in terms of now looking at us having this larger, much more robust data set and applying some of those thoughts differently.

1983
01:59:31.518 --> 01:59:42.527
and learnings to it in addition to continuing to validate and replicate using the plates from the Palomar Sky Survey as well.

1984
01:59:42.768 --> 01:59:52.716
Like there's multiple avenues of exploration here around this idea of these transient light artifacts that are occurring.

1985
01:59:53.797 --> 01:59:58.021
I really was excited to just get your perspective on this.

1986
01:59:58.061 --> 01:59:58.341
Yeah.

1987
01:59:58.601 --> 02:00:02.003
I thought it was a cool set of papers, yeah.

1988
02:00:02.123 --> 02:00:03.825
Definitely the PISP one I liked.

1989
02:00:03.985 --> 02:00:06.106
And just also talking about it from first principles.

1990
02:00:06.446 --> 02:00:14.812
Again, yeah, I think it's important to understand because this is historical data, we're not used to analyzing that, right?

1991
02:00:15.312 --> 02:00:16.233
And the analog...

1992
02:00:17.909 --> 02:00:18.710
Nature of it.

1993
02:00:19.091 --> 02:00:20.132
It's not digital, right?

1994
02:00:20.172 --> 02:00:22.335
You're not sticking a CCD in the back of a telescope.

1995
02:00:22.715 --> 02:00:26.080
This is, at the end of the day, a chemical reaction that's giving us something.

1996
02:00:26.120 --> 02:00:32.067
So it's important to understand the whole chain of custody from photon to pixel.

1997
02:00:32.268 --> 02:00:33.009
Right, right.

1998
02:00:33.069 --> 02:00:34.951
And how do we even get to pixel from photon?

1999
02:00:34.971 --> 02:00:35.131
Yeah.

2000
02:00:36.852 --> 02:00:39.393
an incredible set of stories this week.

2001
02:00:39.433 --> 02:00:40.813
We did our lucky number three.

2002
02:00:41.694 --> 02:00:46.815
We started with our generative AI for cancer detection out of university of Toronto.

2003
02:00:47.335 --> 02:00:50.296
Again, sorry again about the loss, but you're going to save the lives.

2004
02:00:50.316 --> 02:01:01.278
You're going to save the lives of millions of people with, with that, that great, great story followed up by our super story on the retinal wireless retinal chip.

2005
02:01:01.298 --> 02:01:03.159
We're making blind people see again.

2006
02:01:03.179 --> 02:01:03.319
Yeah.

2007
02:01:03.339 --> 02:01:03.519
Yeah.

2008
02:01:04.519 --> 02:01:30.097
not wireless restoring yeah restoring not stopping not slowing down restoring yeah and and we ended with our uh interesting possibility of data related to what have now been renamed uap formerly known as ufos and these transient uh pre-spotnic transients in the palmar sky survey i'm very interested to see people apply some kind of similar

2009
02:01:31.398 --> 02:02:00.043
methodology to the vera rubin data sets they're there they're publicly available you're interested in working on that please shoot us a dm uh we can make the connections uh episode 15 we're continuing to cook uh unbelievable engagement on just i i still can't believe people care about science as much as they do yeah i love to see it it's you do love to see it science is important we are not covering i know people are begging us

2010
02:02:00.823 --> 02:02:28.501
to cover 3i atlas yeah there's no data the government shut down there's no data for us to cover uh however europa europa clipper is in the tail yeah right now we speak right now yeah but we can't do anything yeah because there's nothing to talk about the esa and um the guys around mars have taken photographs of 3i atlas as it went across it's it's reached it's gone through the closest point of approach

2011
02:02:29.322 --> 02:02:31.424
from the sun and now it's on the other side of the sun.

2012
02:02:31.464 --> 02:02:33.966
So we can actually see it again from the earth.

2013
02:02:34.407 --> 02:02:38.470
It's on its way to Jupiter, but we have no data that we can talk about.

2014
02:02:39.031 --> 02:02:41.393
So until we get the data, we won't cover it.

2015
02:02:41.533 --> 02:02:41.974
Yeah.

2016
02:02:42.314 --> 02:02:53.825
I know that Avi Loeb has talked about the non-gravitational acceleration, which can have a prosaic explanation, but we can't know whether it is that prosaic commentary tale that,

2017
02:02:54.874 --> 02:02:55.414
Or not.

2018
02:02:55.594 --> 02:02:58.296
Thingy or not until we get the data.

2019
02:02:58.436 --> 02:02:58.716
Yeah.

2020
02:02:59.036 --> 02:03:00.997
So we hear you guys.

2021
02:03:01.017 --> 02:03:03.358
You want us to cover 3i Atlas.

2022
02:03:03.579 --> 02:03:04.379
Open up the government.

2023
02:03:04.439 --> 02:03:06.100
Give us the data and we will talk about it.

2024
02:03:06.180 --> 02:03:06.980
Yeah.

2025
02:03:07.260 --> 02:03:13.604
I am your host, Lester Nari, joined as always by my co-host and our resident PhD, Krishna Chowdhury.

2026
02:03:14.044 --> 02:03:16.325
This is From First Principles.

2027
02:03:16.825 --> 02:03:17.886
We'll see y'all next week.
