Latest / From First Principles / AI-Generated Genomes, Retinal Implants, and Palomar’s Mystery Lights Explained (EP. 15)

Transcript

AI-Generated Genomes, Retinal Implants, and Palomar’s Mystery Lights Explained (EP. 15) · From First Principles

• 0:00
  Hello, Internet.
• 0:00
  This is your captain speaking, Lester Nare, joined as always by my co-host and our resident PhD, Krishna Chowdhury.
• 0:08
  We have three great stories lined up for you this week, starting with a story out of the University of Toronto,
• 0:15
  which is around generative AI for cancer detection, which is published in Cell Genomics.
• 0:20
  This will be a good one.
• 0:22
  Our second story is a super story of a couple of institutions here involved.
• 0:26
  Science Corporation, Stanford University, UCSF, University of Pittsburgh, and University of Bonn.
• 0:32
  They've created a wireless retinal implant that is helping blind people see.
• 0:38
  This is out of the New England Journal of Medicine.
• 0:40
  And our final third story actually is two papers
• 0:45
  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.
• 1:00
  This is going to be really interesting, covering a lot of ground.
• 1:03
  As always, we're going to have a great time.
• 1:06
  This is From First Principles.
• 1:25
  My friend.
• 1:26
  How's it going?
• 1:27
  Quite well.
• 1:28
  Quite well.
• 1:29
  Back from our travels.
• 1:30
  Well, yeah.
• 1:31
  You know what happened.
• 1:33
  I mean, I do.
• 1:34
  I am.
• 1:36
  You know what happened.
• 1:38
  The Doyers won the World Series.
• 1:41
  Los Angeles brought home another World Championship.
• 1:44
  Back to back, baby.
• 1:45
  Back to back.
• 1:46
  As predicted.
• 1:47
  As predicted.
• 1:48
  We predicted the Chemistry Nobel Prize.
• 1:50
  Yes.
• 1:51
  And now we predicted the World Series.
• 1:52
  Yes.
• 1:53
  I mean, what can we not do?
• 1:55
  What can we not do?
• 1:56
  Yeah.
• 1:58
  LA is obviously just the greatest place to be right now.
• 2:02
  Yeah, yeah.
• 2:03
  I'm still re-watching.
• 2:06
  they're putting out like world series movies and I'm watching that
• 2:13
  A great, great Game 7.
• 2:15
  Yeah, it was an amazing Game 7.
• 2:17
  Amazing series, amazing Game 7.
• 2:19
  You know, shout out to the Jays.
• 2:21
  Put out a good fight.
• 2:22
  Yeah.
• 2:23
  And we're going to start off with a little bit of, you know, deference to Toronto, an incredible city.
• 2:31
  An incredible city, an incredible university.
• 2:34
  The University of Toronto is, you know, one of the great universities in the world of science and technology.
• 2:40
  Yes.
• 2:41
  They're kind of like the...
• 2:45
  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
• 3:16
  And I finally feel like the Yankees did in the 2000s.
• 3:21
  And now I get it because I don't care either.
• 3:25
  I don't care.
• 3:25
  Yeah, whatever.
• 3:26
  But Toronto is a great city and the University of Toronto is a great university.
• 3:32
  Phenomenal.
• 3:33
  Phenomenal research institution.
• 3:35
  Phenomenal institution.
• 3:36
  And we're going to start off with –
• 3:38
  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.
• 3:52
  Yes.
• 3:53
  And so that's sort of the story.
• 3:55
  And we cannot avoid...
• 4:00
  We cannot avoid AI.
• 4:01
  It's becoming much more prevalent.
• 4:03
  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.
• 4:11
  Or chat GPT.
• 4:11
  The G in GPT is generative.
• 4:14
  And so a lot of people think of consumer use cases.
• 4:16
  But in sort of this early stage research arenas, there are some interesting things happening, this being one of them.
• 4:24
  So how exactly are the researchers at the University of Toronto using GenreVe AI for this detection process?
• 4:32
  Right.
• 4:32
  Well, the problem is precision oncology, okay?
• 4:36
  It needs a lot of large genomic data sets.
• 4:39
  Yep.
• 4:40
  But we also want to preserve the privacy of the individuals that are donating that data set, right?
• 4:47
  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...
• 5:04
  Better mitigate that.
• 5:05
  What are the kinds of drugs that we need to actually do this?
• 5:08
  In order to actually train a model to do that job, we need a lot of data.
• 5:15
  But the data has to be, you know...
• 5:20
  without all of these privacy bottlenecks.
• 5:23
  The sort of personal identifiable information.
• 5:27
  Exactly, yeah.
• 5:28
  And that's what's tough, okay?
• 5:30
  And this Oncogan, which is a generative AI pipeline, it was published in...
• 5:36
  I believe cell genomics, cell genomics, Oncogan, it creates a highly realistic synthetic cancer genome just completely within the computer.
• 5:47
  Okay.
• 5:47
  And it bypasses these privacy issues.
• 5:49
  That's actually really interesting.
• 5:50
  Okay.
• 5:50
  Right.
• 5:51
  And the key feature is it mimics all of these computational, complex, mutational landscapes that are characteristic of cancers.
• 6:02
  But it doesn't tie that data with the patient.
• 6:05
  With an individual.
• 6:07
  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.
• 6:23
  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,
• 6:36
  Into all of the regulations.
• 6:38
  Regulation issues.
• 6:38
  Which are there for a good reason.
• 6:40
  Which are there for a good reason.
• 6:41
  So let's go through what precision oncology means.
• 6:45
  Okay.
• 6:46
  Why it's actually hard to actually simulate these genomes.
• 6:50
  Yeah.
• 6:50
  And then what this current paper is doing.
• 6:53
  Okay.
• 6:53
  Okay.
• 6:53
  What this current paper is announcing.
• 6:55
  So the vision of precision oncology is we want to shift from things like chemotherapy and radiation, which is...
• 7:02
  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.
• 7:12
  Okay.
• 7:13
  And we want to make it precision.
• 7:14
  That's, that's where precision comes from.
• 7:16
  Right.
• 7:17
  And so what we want to do is identify mutations and then administer targeted drugs.
• 7:22
  There's a data bottleneck where we have this paradox between sharing and privacy, right?
• 7:27
  Because privacy,
• 7:28
  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?
• 7:41
  But every single genomic identifier is something that is unique to every individual, right?
• 7:47
  You've got 3 billion base pairs.
• 7:48
  And if I know your 3 billion base pairs, I can figure out who you are, right?
• 7:52
  So...
• 7:53
  re-identification is really possible.
• 7:56
  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?
• 8:17
  So there's privacy risks,
• 8:20
  And that can reveal dispositions to disease, potential discrimination.
• 8:24
  And you've got stuff like HIPAA that actually controls and creates barriers for that kind of data sharing.
• 8:30
  It's a very good thing that we're doing that, right?
• 8:32
  Because then without it, you can have potential discrimination on life insurance, long-term care.
• 8:38
  Yeah.
• 8:39
  Right.
• 8:39
  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.
• 8:49
  Yeah.
• 8:49
  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.
• 8:57
  Exactly.
• 8:57
  Yeah.
• 8:57
  You don't want that.
• 8:58
  That's just blatant nonsense.
• 9:01
  Yeah.
• 9:01
  Right.
• 9:02
  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.
• 9:13
  Right.
• 9:14
  So that you can compare one AI model from another.
• 9:18
  Well, with with cancer data sets, it's really hard to do that kind of benchmarking because.
• 9:25
  The real patient samples have sequencing errors.
• 9:29
  You don't actually know what the tumor complexity is.
• 9:32
  There's no actual ground truth.
• 9:34
  So if we were to generate data sets that have a ground truth, then we can easily benchmark all of these other AI models.
• 9:43
  So when someone says, hey, I've got an AI model that I think does better—
• 9:46
  We've actually got a data set where we can test that.
• 9:49
  Yeah.
• 9:49
  Right?
• 9:49
  Yeah.
• 9:50
  So that's a separate issue from the privacy thing.
• 9:53
  But this thing is actually tackling both of them at the same time.
• 9:57
  Interesting.
• 9:57
  And obviously there's a huge economic and human imperative, right?
• 10:00
  Like there's a global cancer burden, 20 million new cases every year, 9.7 million deaths every year.
• 10:07
  There's a huge economic toll.
• 10:10
  Drug development is long and expensive, takes hundreds of millions of dollars.
• 10:15
  And if we have this kind of data set, we can train models that do it way better, right?
• 10:21
  We can limit all of that and democratize.
• 10:25
  This is actually totally an organic anecdote, but literally I was in a work meeting earlier today.
• 10:33
  And unfortunately, one of my coworkers, you know, has a partner that was just diagnosed with stage four.
• 10:41
  And like, you know, I mean, it happens every day.
• 10:44
  Yeah.
• 10:45
  It is present.
• 10:47
  Personally, for millions of people, one, two degrees of separation for so many people.
• 10:51
  Yeah, everyone knows someone.
• 10:52
  And literally today.
• 10:53
  I mean, so this is clearly, obviously, a highly important and prevalent issue.
• 11:02
  Yes, exactly.
• 11:04
  And so let's talk about like cancer itself and why it's even like hard to do this.
• 11:10
  It's hard to like make these genomic data sets that are synthetic the way that we want.
• 11:15
  Okay.
• 11:16
  Cancer is a disease of the genome.
• 11:17
  It's caused by DNA errors.
• 11:19
  At the end of the day, that's what's happening.
• 11:21
  Okay.
• 11:21
  There's 3 billion DNA letters, ATGC, in our genome.
• 11:27
  And
• 11:28
  We're going to use this analogy of the genome being a library of cookbooks.
• 11:32
  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?
• 11:43
  So the genome is a library of these cookbooks.
• 11:46
  Each cookbook is a chromosome, and then each recipe is a gene inside that cookbook.
• 11:51
  Okay.
• 11:52
  Okay?
• 11:53
  If we have errors in critical recipes, that's going to lead to cancer.
• 11:59
  And there's really two types of big mutations that happen, these typos that happen in our recipe.
• 12:06
  Okay?
• 12:07
  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.
• 12:18
  reading of the thing, or I delete a bunch of letters, so on and so forth, right?
• 12:24
  There's two types of mutations that happen.
• 12:26
  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?
• 12:32
  Because that's the, you became, you came from a zygote, which is a single cell, and so whatever mutation was in that zygote,
• 12:39
  That's all of you, right?
• 12:41
  Then there's somatic mutations.
• 12:43
  These are acquired during your lifetime.
• 12:45
  It happens within a single cell.
• 12:47
  And then the progeny of that single cell inherits that somatic mutation, right?
• 12:52
  Cancer is actually primarily driven by accumulated somatic mutations.
• 12:57
  Okay.
• 12:57
  Yeah.
• 13:00
  There are obviously cases where germline mutations create the cancer off the bat.
• 13:05
  Yes.
• 13:05
  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?
• 13:18
  That makes sense.
• 13:19
  One key aspect is Oncogan, which is this generative AI, it's trained on and it generates only somatic mutations.
• 13:27
  It doesn't get trained on the germline info.
• 13:31
  So it's already getting rid of that sort of privacy issue.
• 13:36
  Which is important to understand that there's two entry points, one of which has less of a privacy concern by default.
• 13:44
  And that's where this Oncogan is focused on.
• 13:46
  Exactly, exactly, yeah.
• 13:48
  Now, when it comes to the architects of cancers, there's two types of cancers.
• 13:51
  There's a driver and a passenger.
• 13:53
  So the driver mutation, these are critical alterations that give a growth advantage to some kind of cell.
• 13:59
  Okay?
• 14:00
  It can either be something like...
• 14:02
  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?
• 14:12
  A tumor suppressor.
• 14:14
  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?
• 14:21
  No, not today.
• 14:23
  So these are the driver mutations.
• 14:25
  Then there's passenger mutations, which are kind of like hitchhikers.
• 14:28
  They accumulate by chance.
• 14:30
  There's no growth advantage.
• 14:32
  It's kind of like just there, but it doesn't really affect the...
• 14:37
  the cellular machinery in the way that it like turns it on and becomes cancerous, right?
• 14:42
  Tumors have very few driver mutations amidst thousands of passenger mutations.
• 14:49
  Okay.
• 14:50
  So there's this distribution where there's a few that are targeted for these oncogenes or these tumor suppressor genes that,
• 14:58
  And then there's a bunch of noise.
• 14:59
  Yes.
• 15:00
  Right?
• 15:00
  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.
• 15:10
  Right?
• 15:10
  And Oncogan does exactly that.
• 15:12
  Right?
• 15:13
  It replicates this tumor architecture perfectly.
• 15:15
  by making a bunch of random passengers and not that many driver mutations.
• 15:21
  Okay?
• 15:22
  And the final thing I want to touch on is these mutations have fingerprints when it comes to cancer.
• 15:28
  Okay?
• 15:29
  There's characteristic patterns of mutations that...
• 15:34
  give certain processes a leg up that becomes cancerous.
• 15:39
  For example, like UV light, right?
• 15:40
  UV light causes a C to T mutation.
• 15:44
  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.
• 15:54
  Right?
• 15:54
  Like one leg of your ladder, it's got two legs.
• 15:57
  One leg of your ladder is going to staple itself onto itself and become a dimer.
• 16:02
  And then that's going to cause melanoma, for example, right?
• 16:06
  Tobacco smoke is linked to another single base substitution that'll cause lung cancer.
• 16:13
  So Oncogan, which is this thing that people have made, it accurately reproduces that tissue-specific signature, right?
• 16:22
  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.
• 16:29
  It's actually learned how to do all of that stuff.
• 16:32
  That's fascinating.
• 16:33
  And then the other thing, we're not done with all of the stuff that can cause cancer.
• 16:37
  There's also something called large-scale vandalism, right?
• 16:40
  There's copy number alterations, CNAs.
• 16:44
  This means you've taken a whole chunk of your chromosome
• 16:48
  And you've duplicated it.
• 16:50
  Okay.
• 16:51
  Okay.
• 16:51
  So now there's a duplication or even a deletion sometimes of large chromosomal segments.
• 16:56
  So you've got large parts, like large chapters of your cookbook that are now been copied or completely deleted.
• 17:03
  You've also got structural variants where you take a big chunk of your DNA and you flip it.
• 17:09
  And now it's reversed.
• 17:11
  Now, it's just all sorts of chaos, right?
• 17:13
  Right, right, right.
• 17:14
  Like, there's so many different things that can happen with cancer genomes, right?
• 17:19
  That are the driver and the cause of it.
• 17:21
  Right.
• 17:22
  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?
• 17:35
  Mm-hmm.
• 17:36
  that then can be used to train other AI, to train other tools, right?
• 17:41
  And that's what this paper is doing.
• 17:43
  So they use something called generative AI.
• 17:49
  We've seen this a lot, right?
• 17:50
  Like a cat on a thing.
• 17:52
  This one, this particular architecture is called a GAN.
• 17:56
  It's a generative adversarial network, okay?
• 17:59
  There's two competing neural networks.
• 18:00
  This is a very interesting way to actually train AI and to architecture it up.
• 18:06
  I think it's very cool.
• 18:07
  It's been used actually for image generation as well.
• 18:10
  Yep.
• 18:10
  But this time what they're doing is they're using it to generate genomic data.
• 18:16
  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...
• 18:25
  Now mapping it to this problem set is an interesting use of the architecture of generative adversarial networks.
• 18:32
  That's interesting.
• 18:34
  And the basic idea behind generative adversarial networks is you've got two networks, okay?
• 18:40
  You've got one that's the forger.
• 18:43
  And one that's the discriminator.
• 18:45
  Okay?
• 18:46
  The forger or the generator, this is the guy who's going to create forgeries from random noise and try to attempt realism.
• 18:54
  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.
• 19:06
  So both of these guys have the training set, which is all the stuff that is actually real.
• 19:10
  And the generator is creating fake versions of that.
• 19:14
  Yep.
• 19:15
  And over many, many iterations, both of these guys are going to get really good at their jobs.
• 19:20
  The generator is going to get really good at creating fake data.
• 19:24
  And the...
• 19:26
  discriminator is going to get really good at telling whether it's fake or not.
• 19:31
  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.
• 19:42
  So there's a substrate of truth that these two processes have access to.
• 19:48
  One's job is to create synthetic data based on that substrate.
• 19:53
  Yeah.
• 19:53
  The other is to say whether you're fake or not.
• 19:57
  And because there's this sort of iterative back and forth between these two, you get to a place where –
• 20:06
  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.
• 20:16
  Exactly.
• 20:17
  And now it's like really good at making real looking data.
• 20:21
  Which goes back to our privacy issue and some of the other things we just talked about.
• 20:25
  Exactly.
• 20:25
  Yeah.
• 20:25
  Yeah.
• 20:26
  And so this cycle repeats until the generator produces stuff that's indistinguishable from the real data.
• 20:31
  Yeah.
• 20:32
  And this can be used for tabular data, which is a lot of this genomic data, right?
• 20:36
  It's not like images and things like that.
• 20:38
  It's like tables of like how much of this is there, how much of this is there.
• 20:42
  Oncogan uses something called CTABGAN+, which is just...
• 20:48
  specific architecture that's used for tabular data.
• 20:50
  Got it.
• 20:51
  And it handles mixed data types, imbalanced distributions.
• 20:54
  So like the distribution is not like completely uniform.
• 20:57
  You have some stuff happening all the time.
• 20:59
  Some mutations happen all the time.
• 21:01
  Other mutations are rare, so on and so forth.
• 21:03
  So this takes care of that.
• 21:04
  It handles the complexity of this kind of data set.
• 21:07
  Exactly.
• 21:08
  Exactly.
• 21:09
  Now, the other thing that they're using is something called variational auto encoders.
• 21:13
  These are,
• 21:16
  simpler than GANs.
• 21:17
  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?
• 21:27
  And that latent space has fewer dimensions than the original data.
• 21:34
  And then from that latent space, we have a decoder that tries to recreate the...
• 21:42
  what was originally compressed.
• 21:45
  Yeah.
• 21:45
  Okay?
• 21:46
  Yeah.
• 21:46
  So this is used a lot.
• 21:48
  This is used actually a lot in like image generation.
• 21:51
  When we have, let's say a 256 by 256 image, that has 256 times 256 dimensions, right?
• 22:00
  Because each pixel has a value.
• 22:02
  And you can imagine if there's only three pixels, you can imagine putting that on an XYZ plane, right?
• 22:14
  Like the first pixel is how far it is in the X direction.
• 22:17
  The second pixel is how far it is in the Y direction, and so on and so forth.
• 22:21
  Well, now you have 256 times 256 dimensions.
• 22:26
  Every single image is highly dimensional.
• 22:29
  But if I were to take a bunch of photos of just a bunch of faces, right, the essence of the face is actually...
• 22:38
  not that highly dimensional.
• 22:40
  There's always a nose in the middle.
• 22:42
  There's always a mouth.
• 22:42
  There's always eyes, right?
• 22:45
  Maybe one of the directions could be how dark is the face, right?
• 22:49
  For us, it would be like maxed out.
• 22:52
  Very dark.
• 22:53
  Yeah, very dark.
• 22:54
  Like there would be one dimension that would tell the variational autoencoder, this is a very dark face, right?
• 23:00
  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.
• 23:06
  But...
• 23:08
  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?
• 23:16
  That makes sense.
• 23:16
  The stuff that's varying is the stuff that's going to be encoded in that compressed middle part, right?
• 23:22
  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?
• 23:28
  And puts it, puts it into the actual image that it reconstructs.
• 23:32
  And this is something that we're seeing in MNIST.
• 23:34
  For example, the MNIST dataset is a bunch of handwritten digits.
• 23:38
  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...
• 23:48
  latent space, but I can pick any point in between and that'll result in something real.
• 23:55
  Okay.
• 23:55
  It's continuous.
• 23:56
  It's not discrete like this and this, and there's nothing in between, right?
• 24:01
  There is actually meaning in between.
• 24:04
  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.
• 24:12
  Yeah.
• 24:12
  Yeah.
• 24:13
  to a three, to a six, to a zero, right?
• 24:16
  And it's continuously deforming.
• 24:19
  And all they're doing is reconstructing what happens if I move on that latent space, like as a walk.
• 24:25
  And I see, what is the decoder reconstructing?
• 24:28
  Does that sort of make sense?
• 24:30
  No, 100%.
• 24:32
  It's like, it's taking all of the information of the stuff that it's trained on and saying, what is the essence?
• 24:38
  Yes.
• 24:38
  Of the thing.
• 24:39
  Yes.
• 24:39
  I'm going to put that into a landscape that I can traverse.
• 24:42
  You can traverse.
• 24:43
  Yeah, exactly.
• 24:44
  And then as I move through that landscape, it's going to create meaning with my decoder.
• 24:49
  Because there's different points on that landscape, for example, in the numbers that we find meaning in, which is like the nine...
• 24:56
  The three.
• 24:57
  The zero.
• 24:58
  But there is this transformation.
• 25:02
  Continuous.
• 25:02
  Continuous in between there where you can say, oh, this is like a nine, three-ish thingy.
• 25:10
  Yeah.
• 25:10
  On my way from nine to three.
• 25:12
  Exactly.
• 25:13
  Yeah, exactly.
• 25:15
  Exactly.
• 25:15
  Yeah, yeah.
• 25:16
  So they're using both.
• 25:18
  And the key to using both is that you've got, again...
• 25:22
  that's used for complex interdependent features like mutation counts, driver co-occurrence, that's there due to high fidelity matching.
• 25:32
  And then you have this variational autoencoder that's there for continuous variables, like genomic position.
• 25:38
  The position is a continuous thing, right?
• 25:40
  But the mutation count, that's a discrete thing.
• 25:43
  That's a discrete thing.
• 25:43
  Right?
• 25:44
  So you're combining both into this hybrid approach, right?
• 25:49
  To create this realism in the data set.
• 25:52
  That's actually really, really interesting.
• 25:55
  They're not limiting themselves to one type of architecture.
• 25:59
  Right, right.
• 25:59
  They're really...
• 26:01
  Harnessing the true power.
• 26:03
  It's like there's this blended architecture which is using the value of both of these modalities and taking the best of both.
• 26:10
  Yeah.
• 26:11
  And sort of using them in the context that they're good at in parallel or together.
• 26:19
  Yeah, exactly.
• 26:21
  Exactly.
• 26:21
  Because the genome is both continuous and discrete.
• 26:25
  And so now we're...
• 26:27
  We're getting like two soldiers here that are doing the work, right?
• 26:32
  In order to train Oncogan, they used the training data from Pan Cancer Analysis of Whole Genomes, PCAWG.
• 26:40
  It's a data set that has 2,658 genes.
• 26:43
  whole cancer genomes, like entire genomes, okay?
• 26:47
  All three billion base pairs.
• 26:49
  Massive.
• 26:49
  Massive.
• 26:50
  Massive.
• 26:50
  Yeah, and this data set is treated like sensitive information because it should be, right?
• 26:56
  Because this has actual markers that can identify patients and things like that.
• 27:01
  But using that, they've now created something that will generate data that doesn't have those...
• 27:10
  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.
• 27:21
  Exactly.
• 27:22
  Yeah, yeah, exactly.
• 27:23
  And from what I counted in the paper, there's five GAN models and one TVA model.
• 27:28
  So there's five GAN generative adversarial network models and then one variational autoencoder model.
• 27:36
  all of these guys are working together to create this giant genome, right?
• 27:41
  Multiple giant genomes.
• 27:43
  And obviously the question is to ask is like, you know,
• 27:46
  Okay, how real are you making this, right?
• 27:49
  The photo that you're making, could it fool me?
• 27:52
  This becomes the, with photos, it's easy because we sort of have the eye test, right?
• 27:57
  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.
• 28:03
  I mean, now, dude, I'll be honest, dude.
• 28:06
  There's like videos on Instagram that I get fooled.
• 28:10
  I do a double take and I'm like, oh, this is definitely.
• 28:13
  There's a Neil deGrasse Tyson just did a video of himself talking about how earth is flat.
• 28:19
  And he held up an iPad to the camera.
• 28:21
  You didn't know.
• 28:22
  The video starts with just Neil deGrasse Tyson talking about how the earth is flat.
• 28:25
  And then he just moves the iPad and he's like, that wasn't me.
• 28:30
  And it's.
• 28:31
  It's like, Oh my God.
• 28:32
  And it's the same.
• 28:32
  It's like him in the same space.
• 28:34
  Oh my God.
• 28:35
  The same outfit and the same voice.
• 28:37
  And it's effectively indistinguishable.
• 28:40
  Yeah, dude, it's getting really good, right?
• 28:43
  It's quite good.
• 28:44
  So at least we're harnessing it for something useful.
• 28:46
  Yeah.
• 28:49
  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?
• 28:58
  They looked at mutational density and type.
• 29:00
  They found a high similarity between the synthetic and the real data.
• 29:04
  They looked at mutational signatures.
• 29:06
  They found that it's like...
• 29:08
  pretty nicely correlated.
• 29:10
  The real thing that got me was the genomic distribution.
• 29:13
  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.
• 29:22
  And you can follow where mutations are happening.
• 29:26
  Okay.
• 29:26
  And in cancer, mutations do not happen when,
• 29:30
  uniformly.
• 29:32
  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.
• 29:38
  For example, if you're closer to the center of the chromosome, you're not going to get that many mutations.
• 29:44
  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.
• 29:52
  Right?
• 29:53
  And so they actually showed that mutations in their...
• 29:58
  data set would follow non-uniform distribution.
• 30:02
  So on the top, they have the distribution of mutations in a real sample.
• 30:08
  Yes.
• 30:09
  On the bottom, you've got your generated.
• 30:11
  It looks pretty much the same.
• 30:13
  The eye test.
• 30:14
  The eye test.
• 30:15
  Yeah, exactly.
• 30:17
  And what they've done is like the axis on the x-axis is all the genes.
• 30:21
  Yeah, yeah.
• 30:22
  Like all 23 chromosomes worth of genomes.
• 30:24
  Yeah.
• 30:29
  and they've counted how many mutations are happening, right?
• 30:32
  That's incredible.
• 30:34
  So it's very similar to what we would actually see.
• 30:38
  What we've been basically able to do is now create a system that on the fly, it's not fixed.
• 30:47
  Yeah.
• 30:49
  It's a generative model.
• 30:50
  Yeah.
• 30:51
  So on the fly...
• 30:53
  Any team can now take this and say, we want this kind of dynamics.
• 31:00
  And they can generate now these cancer data sets that mimic real cancer genomes to a degree of –
• 31:11
  accuracy or comparable high fidelity that is usable in practical work in this space.
• 31:20
  Exactly.
• 31:21
  And that's that key that you just touched on, usable, right?
• 31:24
  What does it mean to be usable?
• 31:25
  Well, they actually tried to test it with something called deep tumor.
• 31:30
  Deep tumor is a
• 31:32
  Again, another AI that's been developed to identify cancers.
• 31:36
  Okay?
• 31:37
  They could fool deep tumor really well.
• 31:39
  Oh, wow.
• 31:40
  But here's the key.
• 31:40
  Here's the key.
• 31:41
  Deep tumor is really bad at rare cancer subtypes.
• 31:47
  Okay.
• 31:47
  Okay?
• 31:47
  Like lymph MCLL, there's only 35 samples in the training set.
• 31:52
  Okay?
• 31:53
  And if you have low mutations, you're going to struggle with creating an AI that can identify that.
• 31:58
  Well, they substituted...
• 32:02
  generated samples into the training set, and DeepTumor did better.
• 32:07
  Once deep tumor trained on those, the F1 scores improved.
• 32:11
  That's fascinating.
• 32:12
  The synthetic data from this GAN, this OncoGAN, when fed into an existing detection model.
• 32:20
  That's kind of established.
• 32:21
  People use it all the time.
• 32:22
  It's credible.
• 32:23
  All these things.
• 32:25
  It has now improved its ability for these edge cases.
• 32:27
  These edge niche cases.
• 32:29
  Because now I can generate genomes for that specific thing that's not really prevalent.
• 32:37
  in a lot of the real world data sets.
• 32:39
  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.
• 32:47
  Exactly.
• 32:47
  And then if it comes out with the result, that is really powerful.
• 32:52
  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?
• 33:01
  You do data augmentation where you take whatever data you have
• 33:05
  You like apply transforms to it.
• 33:07
  When it comes to image data, you like squish it.
• 33:09
  You'll like turn it from black and white to red, red, white, and blue or vice versa.
• 33:14
  You'll try to make the model more robust to the data.
• 33:19
  Yeah.
• 33:20
  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
• 33:49
  create usable synthetic data sets that actually map to the real world use cases in a way that is demonstrably true.
• 33:59
  Yes.
• 33:59
  Which is kind of the point, part of the point of this.
• 34:02
  Yeah, exactly, exactly.
• 34:03
  And there's no one-to-one correspondence with the real patients.
• 34:06
  Right.
• 34:06
  You don't have to worry about HIPAA.
• 34:08
  You solve the privacy issue while accelerating the ability to...
• 34:13
  both create the synthetic data sets and then have those applied to existing tool sets or create new tool sets around all this process.
• 34:20
  And the team already, they've got 800 synthetic genomes that are openly available, no ethics approval needed, right?
• 34:27
  So anyone can go and try to train a model based on that.
• 34:31
  If they've got a new architecture in their mind that really is specific to genomic data, they can test it on this.
• 34:38
  It's a new gold standard for benchmarking, right?
• 34:40
  Because now when I generate...
• 34:43
  some kind of data set, I have a ground truth on what type of cancer it is.
• 34:47
  What are the mutations that should be flagged?
• 34:49
  So on and so forth.
• 34:51
  So it's a catalyst for...
• 34:54
  medical oncology, I think.
• 34:56
  I think it's very cool.
• 34:57
  It's a huge new tool in the toolbox for oncology studies.
• 35:01
  Yeah.
• 35:02
  I mean, there's still limitations, right?
• 35:04
  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.
• 35:15
  It's not...
• 35:16
  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?
• 35:32
  It can't simulate complex events like chromothripsis and tumor subclones, but it's still a huge, huge step forward.
• 35:45
  It's one step at a time.
• 35:47
  This has obviously been an issue that is not only affecting so many people, but it's been happening for so long.
• 35:53
  And any new step towards...
• 35:59
  our ability to solve this more robustly is like hugely important.
• 36:02
  Yeah.
• 36:03
  Yeah.
• 36:03
  Yeah.
• 36:03
  Because like cancer is just like, it's just such an insanely difficult problem.
• 36:08
  Right.
• 36:09
  Right.
• 36:09
  It's not a single cause.
• 36:10
  Right.
• 36:12
  There's no single like panacea.
• 36:14
  Right.
• 36:14
  Solution.
• 36:15
  Yeah.
• 36:15
  It's just, yeah.
• 36:17
  So, so any, any incremental, and this, this I think is actually a huge step because I think this opens up
• 36:23
  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.
• 36:34
  Right, right, right.
• 36:35
  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.
• 36:44
  For all the reasons we talked about,
• 36:46
  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
• 37:12
  You know, round three.
• 37:14
  Round three.
• 37:14
  You've given us an incredible new tool in the box.
• 37:17
  Well done on that.
• 37:18
  The ability to do cancer detection and expand the amount of people that can start really working at this issue in earnest.
• 37:26
  That's our story number one.
• 37:27
  Yep.
• 37:28
  Number two, the super story.
• 37:31
  The Avengers came together for this wireless retinal chip.
• 37:37
  Yep.
• 37:39
  That's helping solve for blindness.
• 37:41
  Yeah.
• 37:41
  This is in the New England Journal of Medicine.
• 37:43
  Again, we have Science Corporation, Stanford, UCSF, University of Pittsburgh, and University of Bonn that all collaborated on this story.
• 37:52
  You were really excited to talk to me about this story.
• 37:54
  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
• 38:07
  the retina.
• 38:08
  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.
• 38:19
  and a set of glasses, we can now cure age-related macular degeneration.
• 38:28
  Which is crazy.
• 38:29
  Which is crazy, right?
• 38:30
  The solution is pretty ridiculous.
• 38:33
  Now, to be mindful, this age-related macular degeneration, which is basically age-related blindness in the center of your brain,
• 38:43
  of your visual field.
• 38:45
  This had no prior restorative therapies, right?
• 38:49
  This is the first of its kind that is trying to tackle that.
• 38:53
  Okay.
• 38:53
  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.
• 39:00
  You could see in the peripheral, but in the middle, you couldn't see anything.
• 39:04
  And now we actually have something that is restorative.
• 39:08
  Not just slowing it down.
• 39:10
  Restorative.
• 39:11
  It brings it back.
• 39:14
  The solution is something called the Prima system.
• 39:16
  It's a photovoltaic retina implant microarray.
• 39:19
  It's a novel neuroprosthetic.
• 39:21
  So it's wireless.
• 39:22
  It's subretinal.
• 39:24
  And we're going to get into what all that means.
• 39:26
  This was published in the New England Journal of Medicine.
• 39:30
  The key trial is called Primavera.
• 39:33
  I don't know if the authors are fans of the music festival in Barcelona, but that's what they called it, the Primavera.
• 39:42
  It's an incredible thing, okay?
• 39:48
  81% of participants said they got profound improvement, meaningful improvement, okay?
• 39:54
  And remember those like...
• 39:58
  eye exams that you had to do with like a bunch of big, big like letters.
• 40:03
  And then these guys, the guys who had this over five additional lines on the eye chart.
• 40:09
  Really?
• 40:10
  Right.
• 40:11
  That's, that's, that's a lot.
• 40:13
  I have bad astigmatism and can barely see.
• 40:15
  And so I, I,
• 40:18
  And some patients regain the ability to read letters, numbers, and words.
• 40:23
  That's crazy.
• 40:24
  Okay?
• 40:24
  So let's talk about how the eye works because that's going to tell you about how the mechanism of this thing works.
• 40:31
  Okay?
• 40:31
  The eye is basically a camera.
• 40:33
  Okay?
• 40:35
  You've got a lens.
• 40:37
  You've got...
• 40:38
  a detector in the back, just like a camera.
• 40:40
  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.
• 40:51
  The eye is very, very similar, okay?
• 40:53
  The retina is an image sensor in the back.
• 40:57
  And this retina has photoreceptors, which are your rods and cones.
• 41:02
  You can think of that as your CCD.
• 41:04
  That's the thing that actually interacts with the light.
• 41:06
  And then if...
• 41:07
  a little photon interacts with it, it's going to set off a little electrical signal.
• 41:12
  Then it goes through a bunch of different cells.
• 41:16
  They're called bipolar cells.
• 41:17
  Some of them are called amicron cells.
• 41:20
  And at the end of the day, you've got ganglion cells, which are the cells at the very end that...
• 41:28
  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.
• 41:40
  And that's what's transmitting visual information to the brain's visual cortex, okay?
• 41:45
  But it's really, it's like mechanistically, the eye is remarkably like a camera.
• 41:50
  You've got a lens, and that lens you can squish
• 41:54
  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.
• 42:00
  That's your aperture, right?
• 42:03
  In a normal DSLR, you've got the magnifying, right?
• 42:07
  And the magnification is like how much... Like the distance between your lens and the focal plane.
• 42:17
  Yes, yes.
• 42:18
  But for us, I can focus on the...
• 42:22
  microphone right in front of me.
• 42:24
  And I can also focus on a mountain that's way, way, way far away.
• 42:27
  Right.
• 42:28
  Yeah.
• 42:28
  And the way that our, I can do that.
• 42:30
  We don't have the luxury of actually extending our lens.
• 42:34
  So what we do is we actually change the curvature of the lens lens.
• 42:38
  In order to do so.
• 42:38
  In order to actually get different focal lengths, right?
• 42:41
  I get you.
• 42:42
  So it's remarkably like a camera, okay?
• 42:45
  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.
• 42:56
  It's sort of like the processing center.
• 42:57
  Yes.
• 42:58
  Once the raw data comes in, it kind of packages it up to send it upstairs.
• 43:02
  To send it upstairs, exactly.
• 43:04
  And...
• 43:05
  In the center of your retina, you have something called the macula and the fovea.
• 43:09
  Okay?
• 43:10
  Notice, like, I want you to just do this experiment where you're, like, looking at stuff.
• 43:15
  Notice that you have a lot of detail in the center of your vision compared to the periphery.
• 43:20
  Yeah.
• 43:21
  Right?
• 43:21
  Yeah.
• 43:21
  Like, if you really wanted to read something, like...
• 43:25
  that stuff that's over there, you would focus on it so that it's at the center of your vision.
• 43:31
  If you focus somewhere else, you wouldn't be able to read it.
• 43:34
  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.
• 43:48
  So you've got like...
• 43:51
  Higher resolution there because the pixel density is higher.
• 43:55
  We're not... Everything is not totally in focus all of the time.
• 43:59
  Yes, exactly.
• 44:00
  And macular degeneration is actually a progressive disease that damages that macula.
• 44:06
  So...
• 44:07
  The macula is in the middle, right?
• 44:10
  And you can have two types of AMD, age-related macular degeneration.
• 44:15
  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.
• 44:25
  Those blood vessels can leak, rapid vision loss.
• 44:29
  And then there's dry atrophic AMD, okay?
• 44:33
  And that's the gradual thinning of the macula.
• 44:35
  You get these extracellular deposits where the cells are sort of just dumping their trash, and then that kills everything.
• 44:43
  And all of that accumulation happens, and it kills all of those cones that are there.
• 44:49
  In that center.
• 44:50
  In that center.
• 44:51
  And so that center of your, the center of your visual field Yes.
• 44:57
  is the thing that goes away, which is like, obviously, and that's the whole point.
• 45:01
  Yeah.
• 45:01
  Right.
• 45:02
  Right.
• 45:02
  Right.
• 45:02
  That's, that's where most of my information is coming from.
• 45:06
  Right.
• 45:07
  And that's the one that's more, more common.
• 45:09
  Right.
• 45:09
  Okay.
• 45:09
  And then there's geographic atrophy.
• 45:12
  Um,
• 45:12
  which it's at the end stage of AMD.
• 45:16
  You get these well demarcated patches of your retinal tissue where they die off.
• 45:20
  You get the photoreceptors that are dying off.
• 45:23
  And it results in like absolute scotoma.
• 45:28
  So you get a blind spot right in the middle.
• 45:31
  And when you get the blind spot, you literally can't see in the middle of your visual field, which is so crazy, right?
• 45:37
  It's a complete loss of phototransduction.
• 45:40
  Okay.
• 45:40
  Okay.
• 45:42
  That is what this particular study is trying to handle.
• 45:46
  Okay?
• 45:47
  And it's pretty common, actually.
• 45:49
  Irreversible blindness among older patients globally, you've got something like 200 million people that have this.
• 46:01
  In the U.S. itself, it's a million people.
• 46:04
  It's an incredibly big problem that we need to solve, right?
• 46:08
  Obviously, quality of life goes away.
• 46:11
  There's a loss of independence.
• 46:13
  It's got a devastating impact on all aspects of your life.
• 46:17
  Especially when... Not that this is...
• 46:20
  It's true.
• 46:21
  It's been true forever.
• 46:23
  But particularly when everything is so device-driven.
• 46:26
  And, you know...
• 46:28
  Oh, yeah.
• 46:28
  I didn't even think of that.
• 46:29
  You know what I mean?
• 46:30
  So it's like you have to be able to read and interact with screens.
• 46:35
  And so it's impossible to do so.
• 46:37
  Exactly.
• 46:38
  Especially in a device-driven kind of era.
• 46:41
  Yeah, yeah.
• 46:42
  And, you know, previously it was just like, oh, you're just – you're totally –
• 46:47
  not having a good time.
• 46:49
  Okay.
• 46:49
  There are a few drugs.
• 46:51
  Like there's a, there's the first drug was approved in 2023.
• 46:54
  I survey.
• 46:55
  Um, there's another drug called, uh, syphovray, I guess the mechanism is you basically inhibit this cascade of, uh,
• 47:07
  Stuff that's going wrong in the back of your retina.
• 47:09
  It slows the lesion growth by 20%.
• 47:12
  But at the end of the day, it's not restorative.
• 47:15
  It's just stopping.
• 47:16
  It's just slowing it down.
• 47:17
  The degeneration.
• 47:18
  Right?
• 47:18
  And that's great.
• 47:20
  Yes.
• 47:21
  But for someone who's already had it,
• 47:24
  It's not really doing much.
• 47:25
  Yeah, right, right, right.
• 47:27
  Right?
• 47:27
  And so now we've got this Prima system.
• 47:29
  It's a new paradigm in vision restoration.
• 47:31
  It is restoring things, okay?
• 47:33
  It's got two primary parts.
• 47:35
  There's an internal component.
• 47:37
  There's a tiny, tiny photovoltaic implant, okay?
• 47:41
  It's a little cluster of neuro...
• 47:46
  implants, right?
• 47:47
  It's a tiny little electronic device at the end of the day that goes in the back of your head, okay?
• 47:53
  And it surgically gets placed in the back of your retina, in the back of your eye.
• 47:57
  And then there's an external component, which is specialized glasses that
• 48:02
  with a video camera.
• 48:03
  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.
• 48:23
  That's – okay.
• 48:24
  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.
• 48:35
  But because all the stuff in the macula is now dead, we're not like regenerating the cells itself.
• 48:42
  But we are creating a synthetic version of the exact same input such that when it goes into your brain, it's receiving –
• 48:51
  an identical signal to what you would in a natural sense, but in an artificial...
• 48:57
  Very interesting.
• 48:58
  Exactly.
• 48:58
  It's like simple, but at the same time, extremely hard to do.
• 49:02
  Obviously, because it hasn't been done yet.
• 49:03
  Yeah, and it's only being done because of today's technology.
• 49:06
  So now, let's get into... First, let's tackle the implant.
• 49:09
  What is the thing that is going in the back of your eye?
• 49:13
  It's a 2 nanometer by 2 nanometer little chip.
• 49:17
  30 microns thick.
• 49:19
  It's a silicon chip.
• 49:20
  It's a pixel array of 378 different independent hexagonal chip pixels, two nanometers by two nanometers.
• 49:28
  It's way, way smaller than a penny.
• 49:30
  It's like, it's like, you know, smaller than even Lincoln's.
• 49:34
  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.
• 49:39
  That's how small this thing is, right?
• 49:42
  Small.
• 49:45
  Each of the pixels, there's 378 pixels, each of those pixels has a little mini solar panel and an electrode.
• 49:53
  Now you're thinking, why solar panels?
• 49:55
  Well, solar panels create electricity out of light.
• 50:00
  Now, I've created electricity from the light that is going in.
• 50:06
  Into my eye.
• 50:07
  I don't need an external battery or anything like that.
• 50:10
  I'm so mad.
• 50:12
  I thought that was so cool, dude.
• 50:14
  That's so genius.
• 50:15
  I thought that was so cool.
• 50:17
  Which matters because a lot of times you can even use phones as an example.
• 50:22
  A lot of the weight that goes into the phone is the battery.
• 50:26
  This is the first one that is wireless.
• 50:28
  The reason why it's wireless is because they use this photovoltaic cell.
• 50:32
  That makes sense.
• 50:33
  That's genius.
• 50:34
  And I think that's so genius.
• 50:35
  Yeah, yeah.
• 50:36
  Right?
• 50:36
  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.
• 50:49
  There's a rod and cone there.
• 50:50
  But the energy for that electricity is coming from the photon itself.
• 50:55
  And it makes sense because your eyes are, anyway.
• 50:57
  Right?
• 50:58
  Yeah, that's smart.
• 50:59
  That's so smart.
• 51:00
  That's very smart.
• 51:01
  That's so smart.
• 51:02
  I love that.
• 51:03
  And as soon as I saw it, I was like, oh, that's dope.
• 51:08
  That's dope.
• 51:09
  That's good.
• 51:09
  Yeah.
• 51:09
  And so what they do is they take this little chip, two nanometers by two nanometers, and they put it under the retina.
• 51:15
  Okay.
• 51:15
  This is deliberate where the electrodes are in proximity to the bipolar cells.
• 51:19
  So it's, it's where the, it's where the photoreceptors would be.
• 51:23
  And there you can see on the left, you've got a giant patch of where the photoreceptors have died.
• 51:31
  You're putting that chip right underneath.
• 51:34
  Mm-hmm.
• 51:35
  Where those photoreceptors should be.
• 51:37
  Okay?
• 51:38
  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.
• 51:48
  Right?
• 51:49
  It's natural.
• 51:50
  It's efficiently processed signal.
• 51:52
  And it's more sophisticated than these older retinal implants.
• 51:56
  That we're stimulating the ganglion cells that were directly going into the brain.
• 52:01
  These guys are mimicking the photoreceptors.
• 52:03
  Itself.
• 52:04
  Yeah.
• 52:05
  You see?
• 52:05
  Yeah.
• 52:06
  It's so nice, dude.
• 52:09
  I love that.
• 52:10
  And what I love about this as well is that the pixel has an input.
• 52:17
  So that's where the source of the electric field, like the source of the current.
• 52:21
  But it's also got a sync.
• 52:24
  Because if you don't have a sink, then you put a lot of current out.
• 52:27
  That current is going to spread out and it's going to trigger all the rest of the cells everywhere else in the retina.
• 52:33
  But if you have a sink, then it becomes localized.
• 52:36
  The current goes out.
• 52:37
  It triggers the stuff in its neighborhood and it goes back in.
• 52:40
  It becomes discreet.
• 52:42
  Yes, exactly.
• 52:43
  Exactly.
• 52:44
  So now your pixel size, right, for like what I'm trying to see, that becomes smaller.
• 52:48
  Yeah.
• 52:50
  They've thought of it.
• 52:50
  They've thought of everything, mate.
• 52:51
  This is really good.
• 52:52
  It's really nice.
• 52:53
  This is really good.
• 52:54
  Okay, so that's the implant.
• 52:55
  Okay.
• 52:56
  Now let's get into the outside, right?
• 52:57
  Yes.
• 52:58
  There needs to be something that actually goes and puts light to that implant.
• 53:03
  chip that's in the back of your retina.
• 53:05
  Yes.
• 53:06
  Right?
• 53:06
  And that's where these glasses come in.
• 53:08
  So there's a user interface.
• 53:09
  There's a smart component.
• 53:10
  You put on these glasses.
• 53:12
  You also have a user interface.
• 53:13
  Like it's kind of like a remote.
• 53:15
  And that's your user interface that you can like zoom in.
• 53:20
  Zoom out.
• 53:22
  Yeah.
• 53:22
  You can adjust the contrast and things like that.
• 53:26
  And then your glasses, they've got a little video camera, kind of like the meta glasses.
• 53:31
  You know, the meta glasses have a video camera.
• 53:33
  And those glasses then send in near-infrared.
• 53:39
  signals to the back of your retina, which is where the chip is.
• 53:44
  Why near-infrared?
• 53:45
  Well, all the other natural photoreceptors in your eye are not sensitive to near-infrared.
• 53:52
  Right?
• 53:52
  If you were sending invisible stuff, then you'd be blinded by like a giant...
• 53:58
  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
• 54:14
  It's so nice.
• 54:16
  That's, I'm, I'm.
• 54:17
  Right?
• 54:18
  Yeah, yeah.
• 54:19
  You're seeing why I was like decided about this.
• 54:21
  This is so cool.
• 54:22
  I just, I'm like, I don't have anything to say because it's just, it's so good.
• 54:26
  It's so good.
• 54:26
  Like they thought of it, right?
• 54:28
  It's so good.
• 54:28
  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.
• 54:35
  So you can go into the corner of your eye.
• 54:37
  I mean, it's still a crazy surgery.
• 54:39
  I'll be honest.
• 54:40
  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.
• 54:44
  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?
• 54:56
  And it's actually incredible.
• 55:00
  This chip efficiently absorbs 30 microns of silicon photodiodes, right?
• 55:06
  It's 30 microns of silicon photodiodes.
• 55:10
  You get this pulse light that's coming in at 30 hertz.
• 55:14
  And 30 hertz is, you must know, 30 hertz is about where film frame rate is.
• 55:21
  And that's because the ganglion cells that are going to the back of your head, right, that are relaying this information.
• 55:27
  Anything faster than 30 hertz, and it's really not discernible, anything slower in it would look like stop motion.
• 55:34
  So at 30 hertz, you're getting this flicker-free image of pulse light that's going in.
• 55:40
  This is really clever.
• 55:42
  Bro, it's so cool.
• 55:43
  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?
• 55:52
  There's the actual like where in the eye do you make the intervention in order to get the restorative benefit?
• 55:58
  They made that sort of innovation by putting it in that like where the cones and rods are.
• 56:04
  Yeah.
• 56:04
  Then you have to think about how do you do, how do you get power?
• 56:08
  Yeah.
• 56:08
  That's the solar piece.
• 56:09
  That's the, yeah, you made a solar cell.
• 56:11
  So brilliant.
• 56:12
  Then you have to think about like, how are you actually going to turn it on?
• 56:16
  Meaning like, how are you going to direct the light to make it actually work the way you want it to?
• 56:20
  But not affect everything.
• 56:21
  But not affect everything else.
• 56:22
  And so the near infrared as the methodology to do so is so genius.
• 56:27
  Yeah, dude.
• 56:28
  Oh man, that's so good.
• 56:29
  It's so good.
• 56:30
  That's quite, that's quite nice.
• 56:31
  It's so nice.
• 56:32
  That's quite nice.
• 56:32
  Yeah, it's so nice.
• 56:34
  And let's talk about the study itself that came out in the New England Journal of Medicine, right?
• 56:39
  It's a multi-center.
• 56:40
  So 17 sites across five European countries.
• 56:43
  You mentioned the University of Bonn.
• 56:45
  I mentioned them specifically because they kind of did a lot of the groundwork.
• 56:49
  But a bunch of other European institutions were part of the clinical trial, right?
• 56:55
  Sure, sure, sure.
• 56:55
  But you're increasing generalizability because of these 17 different studies.
• 56:59
  places are trying it.
• 57:01
  Because people culturally, genetically, you want to have it spread across a variety of years.
• 57:05
  Exactly, yeah.
• 57:05
  So 38 participants, all of them over the age of 60, and all of them confirmed atrophy in both eyes.
• 57:12
  Both eyes.
• 57:13
  Right?
• 57:13
  Right.
• 57:14
  And what they looked at was something called the log mar, which is the logarithm of minimal angle of resolution.
• 57:22
  It's effectively saying like in the center of your field of vision, like how, how big can you see things?
• 57:28
  It's this thing.
• 57:29
  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?
• 57:36
  Those letters tell you how small of a letter you can discern.
• 57:40
  What line can you see on this one?
• 57:43
  I mean, the screen is pretty small, but I think I can see that.
• 57:50
  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.
• 58:03
  Highly significant.
• 58:04
  And it's equivalent to five lines.
• 58:07
  Right.
• 58:09
  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.
• 58:23
  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.
• 58:36
  Yeah.
• 58:36
  Yeah.
• 58:37
  permanently is now no longer like necessarily true.
• 58:42
  Yeah, dude.
• 58:42
  There was one dude who went from legal blindness to like, I can drive now.
• 58:47
  That's crazy.
• 58:48
  I mean, that's one dude, but still.
• 58:51
  And this is like state, this is V1.
• 58:53
  Yeah.
• 58:53
  Right.
• 58:53
  Yeah.
• 58:54
  This is V1.
• 58:54
  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.
• 59:08
  Yeah.
• 59:09
  For any number of different permutations or use cases, etc.
• 59:14
  Exactly.
• 59:15
  That's a really, really interesting...
• 59:16
  It's really cool.
• 59:17
  I mean, there were detractors.
• 59:19
  Like, if you looked at user satisfaction, 69%.
• 59:22
  So 22 out of the 32 reported medium to high overall satisfaction, which...
• 59:29
  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.
• 59:35
  And I looked into more of that.
• 59:37
  I think part of that is because of these things called serious adverse events that happen at the end of any surgery.
• 59:42
  Like you're, look, you're sticking a needle into someone's eye.
• 59:46
  or putting in a chip, like there's going to be serious adverse events, right?
• 59:51
  And most of the stuff had to do with the trauma and healing from the surgery.
• 59:55
  Not from the actual device, right?
• 59:58
  So if we can make that process better...
• 1:00:02
  perhaps this can be a lot better.
• 1:00:04
  The satisfaction can be higher.
• 1:00:05
  Yeah, the satisfaction can be higher, right?
• 1:00:07
  Which makes sense.
• 1:00:08
  Yeah.
• 1:00:08
  That does make sense.
• 1:00:09
  But at the end of the day, I think this Prima is a quantum leap in terms of retinal implants.
• 1:00:15
  Before, the first FDA-approved retinal implant was something called the Argus 2.
• 1:00:20
  Okay.
• 1:00:21
  Which had 60 electrodes.
• 1:00:23
  That was this image I just pulled up earlier.
• 1:00:25
  Yeah, and that was wired, though.
• 1:00:27
  Right.
• 1:00:28
  You can see it coming out of the... Yeah, yeah.
• 1:00:30
  And you can see that in the retina itself, there's a little wire that are like...
• 1:00:34
  Right?
• 1:00:35
  This thing is just complete...
• 1:00:36
  It's just a chip naked on its own.
• 1:00:38
  Wireless.
• 1:00:39
  Wireless, dude.
• 1:00:41
  Wireless, no battery.
• 1:00:42
  No battery.
• 1:00:43
  It's just a photovoltaic cell.
• 1:00:46
  It's so clever.
• 1:00:48
  It's so clever, dude.
• 1:00:49
  And what they're doing next, you talked about how this is just the beginning, right?
• 1:00:54
  These pixels are about 100 microns in size.
• 1:00:57
  They're trying to get it down to 20 micron.
• 1:01:01
  20 micron pixels from 100 microns.
• 1:01:04
  So now we'll be able to fit 10,000 pixels instead of the previous 378.
• 1:01:09
  So now I'll be able to get as much fidelity as I do now.
• 1:01:15
  Which is, again, now it's become an engineering problem.
• 1:01:18
  Yes.
• 1:01:18
  Which is much easier to deal with.
• 1:01:19
  Yes, yes.
• 1:01:20
  Like, I'll take an engineering problem all day.
• 1:01:22
  Yes, yes.
• 1:01:23
  20 microns, right?
• 1:01:24
  That's about the size of the rods and cones themselves, right?
• 1:01:29
  You're now starting to...
• 1:01:30
  to get to the limit the biology was getting to.
• 1:01:33
  Obviously, there's going to be challenges, right?
• 1:01:35
  You can't...
• 1:01:35
  Smaller pixel size means that the electric field penetration to the upper layers of that retina is going to be not as good.
• 1:01:43
  So maybe you need these 3D electrode structures, right?
• 1:01:46
  Where you have like pillars...
• 1:01:48
  Right.
• 1:01:49
  Where the pixel is like this and you've got a pillar that injects electricity into the upper layers of the retina.
• 1:01:56
  But again, engineering problem, which which is infinitely more solvable than going from zero to one.
• 1:02:04
  I mean, this thing is like decades in the making.
• 1:02:07
  you know, but it's working and just, it's so clever.
• 1:02:11
  I loved it.
• 1:02:12
  Yeah.
• 1:02:13
  No, that's really, it took a while to get access to the paper.
• 1:02:16
  Oh yeah.
• 1:02:18
  New England journal of medicine.
• 1:02:20
  I don't know why you guys.
• 1:02:21
  Yeah.
• 1:02:22
  I don't, I don't want to pay.
• 1:02:23
  Like, I think it was like a hundred.
• 1:02:24
  I'll be honest.
• 1:02:25
  I think it was like a hundred dollars.
• 1:02:26
  So I asked a friend of mine who, who is a doctor.
• 1:02:29
  Yeah.
• 1:02:30
  Who does research.
• 1:02:31
  Who has reason to access it.
• 1:02:32
  Yeah, I need you to give me a PDF.
• 1:02:34
  We're trying to give you guys some traffic here.
• 1:02:36
  Okay?
• 1:02:37
  We're trying to give some promo.
• 1:02:38
  Yeah.
• 1:02:39
  So, you know, free advertising on the First Principle pod.
• 1:02:45
  We'll do deep dives on the stories.
• 1:02:46
  Yeah, they're like, who are you?
• 1:02:47
  And, like, honestly, give, like, good understanding.
• 1:02:50
  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.
• 1:02:57
  And, like, what the implications are.
• 1:02:58
  And what's, like, the real world...
• 1:03:00
  you know, meaning and the real world meaning for this is like a big deal.
• 1:03:05
  Yeah.
• 1:03:05
  This is a big deal.
• 1:03:06
  Right.
• 1:03:06
  Like now imagine in the future right now, I'm thinking you can, you can bring AI into it.
• 1:03:12
  Of course.
• 1:03:12
  Right.
• 1:03:13
  You could have like the, the, the Iron Man thing where like the thing automatically zooms.
• 1:03:17
  Yeah.
• 1:03:18
  So you don't have to press a button to zoom.
• 1:03:20
  It'll just know what you're looking at and zoom in.
• 1:03:23
  Like, yep.
• 1:03:24
  No, I mean, this kind of like ties to the first story we had around the, how, how,
• 1:03:29
  AI is now being used for cancer detection.
• 1:03:31
  I mean, there's going to be clever ways in which to implement it in this context.
• 1:03:35
  I mean, there's all this stuff.
• 1:03:37
  I mean, these implants and these things, we've talked about Neuralink before on the pod.
• 1:03:42
  Now we're having these retinal implants.
• 1:03:45
  There's a bottleneck around the surgery process and stuff.
• 1:03:48
  We're still meat sacks.
• 1:03:49
  So that's still a bottleneck.
• 1:03:54
  It's still a bottleneck, yeah.
• 1:03:58
  And it's going to be for a while.
• 1:03:59
  For a while.
• 1:04:00
  But we're getting into real interesting places.
• 1:04:03
  I mean, as someone who already has bad vision, you know, my wife has bad vision too.
• 1:04:10
  She actually has – I didn't do LASIK or anything.
• 1:04:12
  I just wear contacts and glasses.
• 1:04:15
  But her vision was so bad that at the time, like, LASIK wasn't good enough for her eyes.
• 1:04:19
  So she got the implanted contact lens.
• 1:04:22
  Oh.
• 1:04:23
  Like in her eyes.
• 1:04:24
  Wow.
• 1:04:25
  Which is like – and I was like –
• 1:04:27
  That's crazy that you can do that.
• 1:04:29
  And this was like 15 years, you know, whatever ago.
• 1:04:31
  This was a while ago.
• 1:04:33
  And at the time, it was state of the art and this and that and yada yada.
• 1:04:37
  But, I mean, this is like...
• 1:04:39
  orders of magnitude more difficult than just embedding a contact lens in the eye.
• 1:04:47
  Fascinating, fascinating, fascinating.
• 1:04:48
  Yeah, I thought that was such a cool story.
• 1:04:50
  That's a great story.
• 1:04:51
  That is a great, great story.
• 1:04:54
  It gives me hope that if I start getting blind later in life, there's still a chance.
• 1:04:58
  Yeah, there's still a chance.
• 1:04:59
  There's still a chance.
• 1:05:00
  So you're saying there's a chance.
• 1:05:01
  You're saying there's a chance.
• 1:05:03
  We're going to end with our story number three, which is our story about...
• 1:05:08
  or related to at least unidentified anomalous phenomenon, which is more commonly known as UFOs.
• 1:05:13
  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
• 1:05:35
  events that have been happening, particularly within the apparatus of the U.S. government, intelligence, and military-industrial complex structures.
• 1:05:45
  And, you know, this is a subject I've talked to you about a lot.
• 1:05:48
  Yeah, and you're very passionate about it.
• 1:05:50
  It's something that I have interest in.
• 1:05:53
  I work at a nonprofit that's working in this space, the Disclosure Foundation, that's
• 1:05:58
  And one of the things that people always get caught up on is origin.
• 1:06:01
  And what I would ask is that we sort of keep origin to the side for a second in this conversation.
• 1:06:08
  And just talk about what is without trying to ascribe an origin to it.
• 1:06:12
  Because these are all things that are happening regardless of what you want to ascribe the origin to.
• 1:06:18
  And it's interesting regardless of what the origin is.
• 1:06:23
  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.
• 1:06:35
  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.
• 1:06:48
  And that was sort of the catalyst for the modern era.
• 1:06:53
  Prior to that, that is outside of the scope of this episode.
• 1:06:57
  Yeah.
• 1:06:58
  And so we're not going to touch on sort of the, you know, the what I call the legacy era of the subject.
• 1:07:07
  But after that New York Times article came out in 2017, it contained –
• 1:07:12
  three videos.
• 1:07:13
  This was the first time we were getting videos with chain of custody from the U S government that were showing these anomalous objects.
• 1:07:20
  Right.
• 1:07:20
  Okay.
• 1:07:21
  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.
• 1:07:32
  That's changed since that time period.
• 1:07:34
  But the Pentagon actually confirmed that the, what are colloquially known as the gimbal,
• 1:07:42
  are legitimate videos that have chain of custody to the US government.
• 1:07:45
  That happened between 2017 and 2019.
• 1:07:49
  What sort of proceeded from there was an escalation of the...
• 1:07:56
  issue from a national security perspective.
• 1:07:59
  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
• 1:08:23
  What are, what are these UAP?
• 1:08:24
  Where is this unidentified anomalous phenomena?
• 1:08:26
  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.
• 1:08:51
  Right.
• 1:08:51
  Again,
• 1:08:52
  independent of origin or any of these other issues.
• 1:08:55
  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,
• 1:09:12
  Are you all seeing this?
• 1:09:14
  Where are you seeing this stuff?
• 1:09:16
  What is going on here?
• 1:09:18
  That eventually now rolled into what we now know as this permanent Pentagon office that's called the All Domain Anomaly Resolution Office.
• 1:09:29
  That was established in 2022.
• 1:09:32
  And that office was congressionally mandated to be created after that initial report came out.
• 1:09:39
  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.
• 1:09:53
  And it's important to note that the term UAP is a specific term of art.
• 1:09:58
  that does not include any other form of unmanned aerial systems.
• 1:10:03
  So this is drones, this is undersea, unmanned.
• 1:10:08
  Those are all discrete and defined, and there's processes we have to deal with that.
• 1:10:14
  The reason why this became an issue within the government is that these platforms were exhibiting
• 1:10:22
  like characteristics or behaviors that were beyond the scope of what was deemed as current or even next generation technologies.
• 1:10:32
  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,
• 1:10:44
  that was proposed initially in 2023, but then passed in 2024, that mandated that executive agencies release information around the history of the subject.
• 1:10:55
  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
• 1:11:13
  has been an ongoing issue that Congress has been going after and the executive branch has been unwilling to provide information on.
• 1:11:21
  So this is sort of a live issue.
• 1:11:23
  NASA had a UAP independent study group.
• 1:11:26
  They put out their own report around this issue.
• 1:11:28
  Um,
• 1:11:30
  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.
• 1:11:38
  There's been multiple whistleblower protections that have been passed.
• 1:11:41
  There's been defunding of UAP programs for the intelligence community.
• 1:11:45
  I mean, there's this long list of...
• 1:11:47
  of indirect evidence around this subject.
• 1:11:52
  But there's been a dearth of scientific research.
• 1:11:55
  Yes.
• 1:11:56
  Really the main primary thing that has existed in this space, particularly in the last couple of years, have been research papers...
• 1:12:03
  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.
• 1:12:13
  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.
• 1:12:23
  But there are two papers that have just come out that are a first of its kind in terms of having actual data
• 1:12:33
  that again can be potentially either reproduced or replicated in other arenas from the Palomar Sky Survey.
• 1:12:41
  And we're going to kind of go in from a first principles perspective to understand what these papers actually were talking about.
• 1:12:49
  Yeah.
• 1:12:49
  Yeah.
• 1:12:50
  So there's a lead author.
• 1:12:52
  Dr. Beatrice Villarreal.
• 1:12:55
  Villarreal, yes.
• 1:12:56
  Villarreal.
• 1:12:56
  Yes.
• 1:12:57
  I hope I'm saying that right.
• 1:12:58
  Yes.
• 1:12:58
  So the lead author, Dr. Beatrice Villarreal, she actually covered this issue before in her 2021 paper.
• 1:13:06
  Yes.
• 1:13:07
  That was published in Nature Scientific Reports.
• 1:13:09
  It was on nine simultaneous transients occurred in April 1950.
• 1:13:14
  Yeah.
• 1:13:16
  And it's important to note that research on this has been done.
• 1:13:21
  Has been done.
• 1:13:22
  And this is kind of a follow-up paper, but it's, I believe, a lot more thorough.
• 1:13:28
  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.
• 1:13:38
  Sure.
• 1:13:38
  But...
• 1:13:39
  You know, on this podcast, we always focus on from first principles, what is the science that you did?
• 1:13:45
  And does it warrant the conclusions that you put out?
• 1:13:52
  Yes.
• 1:13:52
  Right?
• 1:13:53
  So we're going to start with these papers.
• 1:13:55
  The papers focus on the Palomar Observatory Sky Survey.
• 1:13:59
  Okay?
• 1:14:01
  This is POSS1.
• 1:14:04
  It was a sky survey that was taken up by the Samuel Ostgen Telescope.
• 1:14:09
  which is a camera that is mounted on the Samuel Oshin Telescope at Palomar Observatory.
• 1:14:13
  There you can see the telescope itself.
• 1:14:15
  It's got a Schmidt camera.
• 1:14:16
  And this thing is specialized for wide field photography.
• 1:14:20
  These cameras are actually pretty insane because...
• 1:14:23
  So...
• 1:14:24
  That camera, the camera that is attached to it, has a field of view of 6.6 degrees by 6.6 degrees.
• 1:14:32
  Okay, so 6.6 degrees by 6.6 degrees.
• 1:14:35
  The full moon, for comparison, is half a degree.
• 1:14:39
  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.
• 1:14:46
  This is 13 full moons by 13 full moons.
• 1:14:47
  So it's 170 full moons.
• 1:14:49
  Mm-hmm.
• 1:14:53
  Big is the field of view of this camera, right?
• 1:14:56
  And because it's so big, there's a specialized camera called a Schmidt camera.
• 1:15:02
  And what it does is you've got a primary mirror, right?
• 1:15:05
  The mirror is about 1.2 meters.
• 1:15:07
  That mirror then goes into the camera.
• 1:15:10
  You can imagine the light is getting warped, right?
• 1:15:15
  As the light comes in, it gets warped.
• 1:15:17
  And so the edges of your field of view are going to get distorted.
• 1:15:21
  Yep.
• 1:15:22
  Yep.
• 1:15:23
  This specific camera is specialized for that kind of wide field photography.
• 1:15:28
  So it corrects for that distortion.
• 1:15:30
  And even then, the focal plane...
• 1:15:34
  is actually not flat.
• 1:15:35
  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.
• 1:15:45
  Here, because it's such a large field of view and it's so big, the focal...
• 1:15:51
  the focal plane is actually curved.
• 1:15:53
  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.
• 1:16:04
  And so you got to like warp it before you slide it in there and get this photograph.
• 1:16:09
  I thought that was really cool.
• 1:16:10
  That's interesting.
• 1:16:10
  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?
• 1:16:18
  Given the technology ahead of time.
• 1:16:19
  Now, this is pretty...
• 1:16:19
  Purely photographic, which means that the other cool thing is that the telescope that we just saw...
• 1:16:28
  That telescope is an equatorial mount.
• 1:16:31
  And what that means is that one of the axes is parallel to the Earth's axis of rotation.
• 1:16:35
  And so all you have to do is rotate along that axis, and it'll follow the stars as the Earth rotates underneath.
• 1:16:43
  The stars are all going to move in a circle around the North Star.
• 1:16:47
  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.
• 1:16:55
  in the celestial sphere.
• 1:16:57
  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.
• 1:17:16
  So there was a lot of like manual stuff going on back in the day.
• 1:17:19
  And I think this is really important to see, to, to understand that,
• 1:17:23
  how the data was collected, right?
• 1:17:25
  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.
• 1:17:45
  So this guy's survey produced a thousand photographic plates
• 1:17:50
  About 1,000.
• 1:17:51
  One in red, one in blue, so then you'd have about 2,000.
• 1:17:55
  This is an example of one of the photographic plates.
• 1:17:58
  It's a one degree by one degree patch of the plate that is targeting Pleiades, the seven sisters.
• 1:18:05
  And the haziness that you see, that's actually not a defect.
• 1:18:10
  Those are actual clouds around those stars, those dust clouds around those stars that are creating that.
• 1:18:16
  Right.
• 1:18:17
  It's a beautiful, beautiful example of one of the photographic plates that was actually taken during that time.
• 1:18:23
  The typical exposure was about 45 to 50 minutes.
• 1:18:27
  What we're focusing on here are the red-sensitive plates.
• 1:18:30
  They used to take red-sensitive and blue-sensitive.
• 1:18:33
  Here we're focusing on the red-sensitive because that's what was used in the study here.
• 1:18:37
  And the limiting magnitude is 22 in terms of apparent brightness.
• 1:18:42
  The higher the magnitude, the dimmer.
• 1:18:45
  the thingy is.
• 1:18:46
  And this is about 1 million times fainter than what the human eye can see.
• 1:18:50
  Okay.
• 1:18:50
  So that's what's limiting the faintest object that this telescope can see.
• 1:18:55
  Can see.
• 1:18:56
  Okay.
• 1:18:56
  That's like the lower bound.
• 1:18:57
  Yes, that is the lower bound.
• 1:18:59
  So that's the telescope that we're using.
• 1:19:00
  Okay.
• 1:19:00
  And that's how we're getting these images.
• 1:19:02
  Okay.
• 1:19:03
  Now, the photographic plates.
• 1:19:06
  Okay.
• 1:19:06
  We got to talk about the photographic plates.
• 1:19:08
  We got to talk about the plates.
• 1:19:09
  What does that mean?
• 1:19:10
  Yes.
• 1:19:10
  So photographic plates are 14 by 14 inches.
• 1:19:14
  And these 14 by 14 inches are placed at the back of the camera where all of the light is coming in.
• 1:19:20
  It's one millimeter thick.
• 1:19:22
  These are incredibly thin for how big they are.
• 1:19:24
  They're like the size of a record, you know, a vinyl record.
• 1:19:27
  Yeah.
• 1:19:30
  They've got a photographic emulsion on them, okay?
• 1:19:34
  A light-sensitive silver halide crystal layer.
• 1:19:37
  Usually it's silver bromide, and that's suspended in gelatin, okay?
• 1:19:43
  Bromium is the halogen that they're using here.
• 1:19:46
  Halogen is like that column on the periodic table that's second to last on the right.
• 1:19:50
  Okay.
• 1:19:51
  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.
• 1:19:59
  And then the process is a photon comes in from your telescope and it liberates an electron in the silver bromide.
• 1:20:06
  And that electron, that liberation of the electron forms a latent image spec.
• 1:20:12
  You've got this free...
• 1:20:14
  silver atom, okay, that creates a silver atom cluster.
• 1:20:17
  Then during development, that latent image spec catalyzes, and the silver bromide that's around there goes and turns into black metallic silver, okay?
• 1:20:30
  And then finally you have a washer, a kind of fixer that removes any unexposed silver bromide.
• 1:20:36
  This is like the old school, you know, where you like wash the thingy.
• 1:20:40
  And then that removes any silver bromide.
• 1:20:43
  And what you're left with is a permanent negative image.
• 1:20:46
  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.
• 1:20:56
  Okay?
• 1:20:56
  Right.
• 1:20:56
  Right.
• 1:20:57
  Because that's where the silver is getting deposited.
• 1:21:00
  Yes.
• 1:21:00
  Okay?
• 1:21:00
  So you get this negative image.
• 1:21:02
  Yeah.
• 1:21:03
  Right.
• 1:21:03
  Of the dark is actually where the light is coming in.
• 1:21:06
  Got it.
• 1:21:06
  Okay?
• 1:21:07
  And the white is where there was no photo reaction that happened.
• 1:21:11
  Okay?
• 1:21:11
  And so that's how we actually get these photographic plates.
• 1:21:16
  That's what a photographic plate is.
• 1:21:17
  That's what a photographic plate is.
• 1:21:20
  And that is what the chemical reaction of what a photographic plate means.
• 1:21:25
  Okay?
• 1:21:26
  Now, what this...
• 1:21:28
  What this sky survey is doing and the data that this particular group used is the digitized version of those photographic plates.
• 1:21:40
  So the digitized version in the 1980s and 1990s, we did a digitization process where those glass plates used micro densitometers for...
• 1:21:51
  that basically went through and digitized these photographic plates.
• 1:21:54
  You've got a light beam that measures the opacity of the photographic plate, and that converts that opacity into a digital pixel value.
• 1:22:03
  So the more opaque it is, that means that there was more dark, right?
• 1:22:08
  There was more darkness.
• 1:22:09
  And then if there's more darkness, that means that there's more light over there.
• 1:22:16
  And so now you've converted from...
• 1:22:21
  Photographic plates to a digital sort of pixel values.
• 1:22:25
  Yes.
• 1:22:25
  Right?
• 1:22:25
  Yes.
• 1:22:26
  And there were two digitization processes that happened.
• 1:22:30
  Okay.
• 1:22:30
  There was the digital sky survey that was done with a pixel size of 25 microns.
• 1:22:37
  Okay.
• 1:22:37
  And then there was a super cosmos sky survey.
• 1:22:40
  Yep.
• 1:22:40
  And that was done at the Royal Observatory of Edinburgh.
• 1:22:46
  With a higher resolution of 0.7 arc seconds per pixel.
• 1:22:52
  Got it.
• 1:22:52
  Okay.
• 1:22:53
  So the first one was 1.7 arc seconds per pixel.
• 1:22:56
  And then this one's 0.7 arc seconds.
• 1:22:58
  So it's just like more grainier detail.
• 1:23:02
  Yeah.
• 1:23:03
  You know?
• 1:23:03
  Sort of a thing about.
• 1:23:05
  Slightly higher fidelity.
• 1:23:06
  Yes.
• 1:23:07
  Exactly.
• 1:23:08
  And now you've got a vetting step, right?
• 1:23:09
  Where now...
• 1:23:12
  We can create computer algorithms that look at these digital photographic versions and then try to find transients.
• 1:23:21
  Little spots in the sky that aren't there in current photographs of that patch of the sky.
• 1:23:30
  But that were there.
• 1:23:31
  That were there back then.
• 1:23:33
  And what we can also do is we can rule out any scanning artifacts because we've got two different digital versions that...
• 1:23:40
  Of the photographic plate.
• 1:23:41
  So if something happened in the scanning where I got a little spec, that shouldn't be there in both at the same spot.
• 1:23:47
  Yeah, yeah, right.
• 1:23:48
  The probability of that is insane.
• 1:23:49
  It's insanely high.
• 1:23:51
  The point is the digital scans, there's two versions.
• 1:23:57
  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.
• 1:24:07
  In such a precise way.
• 1:24:10
  And so what you're sort of...
• 1:24:13
  There's a layer of validation.
• 1:24:16
  Yeah, there's like a redundancy here.
• 1:24:17
  A redundancy here.
• 1:24:18
  Because we have two digital... Yeah, we have two digital versions of this.
• 1:24:22
  Okay.
• 1:24:22
  And so...
• 1:24:24
  Now we get into the Vasco project, right, which is the vanishing and appearing sources during a century of observations.
• 1:24:30
  This is led by Beatrice Villarreal.
• 1:24:34
  And the point of this project is to have a systematic comparison between the digitized versions and modern sky surveys.
• 1:24:44
  You look at what the Palomar Observatory saw back then, and you look at what we're looking at right now.
• 1:24:50
  And you've got an automated software that flags these vanishing, the present then, missing now, right?
• 1:24:57
  And there's also appearing.
• 1:24:59
  So not present then, but appearing now, right?
• 1:25:03
  You've got an automated software that flags these things.
• 1:25:06
  And initially, you've got a catalog.
• 1:25:08
  of 100,000 transient candidates.
• 1:25:12
  And here we're seeing two different images, one from POS1, which is the first sky survey, and POS2.
• 1:25:18
  And the circles represent...
• 1:25:20
  two little spots that you can see are exactly not there in the second one, right?
• 1:25:26
  It's like pretty obvious that those two dots were there before and now they're not there now, right?
• 1:25:31
  And so that's a transient.
• 1:25:32
  You've got a point-like source that's there in some plates, but in the later plates, it's not there.
• 1:25:38
  And the idea is, you know, we have two points in time.
• 1:25:41
  And we should be seeing identical things in both those points in time.
• 1:25:44
  Yes.
• 1:25:45
  If it's a star, then yeah.
• 1:25:46
  Because of... And it's not identical.
• 1:25:49
  It's not identical.
• 1:25:50
  Right?
• 1:25:51
  So, people are going to be skeptical about this, right?
• 1:25:57
  Right?
• 1:25:58
  For example, it could just be plate defects.
• 1:26:01
  Sure.
• 1:26:02
  Right?
• 1:26:02
  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.
• 1:26:12
  Well, that means it could just be in the plate.
• 1:26:15
  itself.
• 1:26:16
  What is a hallmark of something that is in the plate?
• 1:26:19
  And this is something that comes from Nigel Hambly at the University of Edinburgh.
• 1:26:23
  He's one of the critics.
• 1:26:24
  He looks at something called the full width at half maximum of these points of light.
• 1:26:30
  So the full width at half max is basically, imagine you've got a Gaussian.
• 1:26:36
  What you're looking for is what is the width of that Gaussian?
• 1:26:39
  When the height is half the peak height, okay?
• 1:26:44
  And what this tells you is something about the noise in the data gathering process.
• 1:26:50
  Now, I told you this telescope, right?
• 1:26:56
  It's traversing the night sky.
• 1:26:58
  Yes.
• 1:26:59
  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?
• 1:27:06
  So even a stationary point of light is going to have a little bit of jitter
• 1:27:11
  As I move this telescope, the telescope is going to not totally track it completely.
• 1:27:16
  And even a point of light like a star that is definitely stationary over the 50 minutes at least.
• 1:27:23
  It might be moving over several years, but over 50 minutes it's stationary.
• 1:27:27
  That's going to have a little spread in my photographic plate.
• 1:27:31
  And what he noticed was these transients, the ones that sort of appear and disappear,
• 1:27:36
  They've got a spread that's way smaller than the stars.
• 1:27:43
  Okay?
• 1:27:44
  Yeah.
• 1:27:45
  And that, to him, was a telltale sign that this is a defect.
• 1:27:48
  Because if it was a light source, then the light source should have the same sort of error that the stars do.
• 1:27:55
  Because they're all light.
• 1:27:56
  Yeah.
• 1:27:56
  Right?
• 1:27:57
  Does that make sense?
• 1:27:57
  It does.
• 1:27:58
  It does.
• 1:27:58
  Right?
• 1:27:58
  Like all of the other things have the same sort of spread because it's from the same jitter of the telescope.
• 1:28:03
  It's uniform across the system.
• 1:28:05
  Right?
• 1:28:05
  Yeah, yeah, yeah.
• 1:28:06
  It's a key thing.
• 1:28:07
  Yeah, yeah.
• 1:28:08
  Now, the contention is, the sort of counter argument is, well, if it's truly transient, meaning like it only like lit up.
• 1:28:17
  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.
• 1:28:29
  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?
• 1:28:38
  But now we're here, we're at a degeneracy problem, right?
• 1:28:41
  Because a genuinely short-lived astronomical event
• 1:28:46
  something that happens for a fraction of a second, is going to look identical to a plate artifact.
• 1:28:52
  Got it.
• 1:28:54
  It's like you're looking for a short-lived astronomical event, but that's going to look exact.
• 1:29:01
  How do you distinguish between what is being argued as a plate artifact versus a true short-lived transient event?
• 1:29:10
  Because it's hard to distinguish between those two.
• 1:29:14
  Yeah, yeah, yeah.
• 1:29:15
  The measure that we're using is degenerate to discerning between one or the other.
• 1:29:22
  Does that make sense?
• 1:29:24
  It's one of the two.
• 1:29:25
  Yeah.
• 1:29:26
  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.
• 1:29:30
  Got it.
• 1:29:31
  Right?
• 1:29:32
  So which one is it?
• 1:29:34
  And that's why there was a necessity for an independent line of evidence.
• 1:29:38
  And this is where we're getting into the two papers that we have now.
• 1:29:41
  Okay.
• 1:29:42
  So it's this challenge of, well, it could be a defect, or it could be a truly transient object.
• 1:29:49
  Which one is it?
• 1:29:50
  And the interesting note here, just as a caveat or a point of order, is there was an emphasis on this being, you know,
• 1:30:01
  The Polymar Sky Survey, which happened in the 50s, was pre-Sputnik.
• 1:30:06
  Yes.
• 1:30:07
  Sky Survey.
• 1:30:09
  And nowadays, when we talk about transient light artifacts, it's...
• 1:30:17
  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
• 1:30:41
  distinguishing between whether it's a plate artifact or not, is if it's not, we're seeing these transient light anomalies that,
• 1:30:52
  At a time period where we... We at least didn't put anything up there, right?
• 1:30:57
  On purpose.
• 1:30:58
  On purpose.
• 1:30:58
  Right?
• 1:30:59
  This is from November 1949 to December 1958.
• 1:31:03
  This is before Sputnik.
• 1:31:05
  Which could be interesting.
• 1:31:06
  Yes.
• 1:31:06
  Because of the fact that it can't just be pointed to being Starlink passing by.
• 1:31:12
  Exactly.
• 1:31:13
  Exactly.
• 1:31:14
  And so now that is where we're getting into the two papers that have come out from the Vasco project, okay?
• 1:31:18
  Okay.
• 1:31:19
  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...
• 1:31:37
  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.
• 1:31:53
  And then there's a second paper that talks about temporal correlations in time between the...
• 1:32:00
  Stuff that we see in the photographic plates and stuff like nuclear tests and UAP sightings, which we'll get to.
• 1:32:07
  So, study one.
• 1:32:09
  This is the alignment hypothesis.
• 1:32:11
  This is Villa Royale and others in 2025 in PASP.
• 1:32:14
  Yes.
• 1:32:14
  Okay.
• 1:32:18
  The paper is aligned multiple transient events in the first Palomar sky survey.
• 1:32:22
  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.
• 1:32:31
  And they could be real physical objects.
• 1:32:33
  And that is the argument that they're making.
• 1:32:35
  Okay.
• 1:32:36
  So they've got a catalog of about 300,000 short duration transients from this Palomar survey that we just went into.
• 1:32:43
  Yes.
• 1:32:43
  Okay.
• 1:32:44
  Okay.
• 1:32:45
  And the first thing that they do is they ask how many of these things are aligned.
• 1:32:52
  They're in a line.
• 1:32:53
  Okay.
• 1:32:54
  And here we've got a photo of plate five.
• 1:32:58
  This is candidate five.
• 1:33:00
  There's five different thingies.
• 1:33:02
  that are all in a line, okay?
• 1:33:05
  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.
• 1:33:16
  Okay.
• 1:33:17
  Okay?
• 1:33:18
  And they find a sigma of 3.9
• 1:33:21
  That this is not chance.
• 1:33:23
  Okay?
• 1:33:24
  3.9 is...
• 1:33:25
  It's pretty high.
• 1:33:26
  Okay?
• 1:33:27
  It's not like 5 Sigma.
• 1:33:28
  Which we've talked about in our previous episode about the strongest evidence yet for life on an exoplanet, which was 3 Sigma.
• 1:33:39
  Which was 3 Sigma.
• 1:33:40
  It was 3.9.
• 1:33:40
  It's higher, but it's still not 5 Sigma.
• 1:33:42
  Sure.
• 1:33:43
  Right?
• 1:33:43
  But it's...
• 1:33:45
  Kind of interesting.
• 1:33:46
  Okay?
• 1:33:47
  And this is just candidate five.
• 1:33:48
  There's other candidates that also have around three sigma, right?
• 1:33:52
  And they go through this in their paper.
• 1:33:56
  That's actually not personally what I found the most interesting.
• 1:33:59
  The alignment is like...
• 1:34:01
  Like, I don't know, by chance.
• 1:34:02
  I don't really know how to reason about that in my head, about how often to do things by chance.
• 1:34:08
  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?
• 1:34:22
  Right.
• 1:34:23
  The Earth's shadow test is, I think, a decisive diagnostic.
• 1:34:27
  So here's the logic.
• 1:34:28
  You've got the Sun, you've got the Earth, and the Earth casts a shadow on the rest of space, right?
• 1:34:34
  When the Moon goes in the Earth's shadow, that's how you get a lunar eclipse.
• 1:34:38
  So...
• 1:34:41
  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...
• 1:34:55
  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.
• 1:35:03
  Right.
• 1:35:03
  Because it's not dependent on whether the thingy...
• 1:35:06
  Right.
• 1:35:06
  Whether the patch of sky I'm looking at is in Earth's shadow or not.
• 1:35:10
  Right.
• 1:35:11
  If, on the other hand, these objects are reflecting the light from the sun, then there should be fewer transients...
• 1:35:20
  When I'm looking in Earth's shadow than when I'm not.
• 1:35:25
  Right.
• 1:35:25
  Okay?
• 1:35:26
  And I know the coordinate of the spot in the sky where I'm looking, right?
• 1:35:34
  Because each of these photographic plates has a right ascension and declination, which is our latitude and longitude in the sky.
• 1:35:40
  We have our time of day.
• 1:35:43
  So from that, I can calculate which patch of the photographic plate is in Earth's shadow and which patch is not.
• 1:35:49
  I do have to do an assumption based on how high I think these objects are.
• 1:35:54
  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.
• 1:36:07
  And at that point...
• 1:36:10
  I can calculate which part of my photographic plate is going to be in its shadow and which part is not.
• 1:36:18
  And then from that, I get an expected number of 1,200 objects, but I only observe 349.
• 1:36:27
  So that's a pretty statistically significant sample.
• 1:36:31
  And this is where I remember when you first came to me with this story, you said 22 sigma significance, right?
• 1:36:39
  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
• 1:37:03
  in Palomar didn't uniformly sample the entire night sky.
• 1:37:08
  The 22 Sigma comes from the assumption that the photographic plates are completely uniformly sampling the night sky.
• 1:37:15
  On the other hand, some patches were looked at more than others, right?
• 1:37:18
  And some patches were completely missed.
• 1:37:20
  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.
• 1:37:29
  How often would I see defects versus not defects and all that other kind of stuff?
• 1:37:34
  And there they get a lower significance, but it's still a lot.
• 1:37:36
  7.6 sigma.
• 1:37:39
  That sits better with me.
• 1:37:41
  Okay?
• 1:37:41
  7.6?
• 1:37:41
  Yeah.
• 1:37:42
  Seems...
• 1:37:45
  more reasonable.
• 1:37:46
  And, and the benchmarking that they did kind of made more sense to me, right?
• 1:37:49
  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.
• 1:37:58
  Well, it didn't.
• 1:37:59
  Right.
• 1:37:59
  It's not the Vera Rubin observatory.
• 1:38:01
  Right.
• 1:38:01
  Right.
• 1:38:02
  It's not just like, yeah, yeah.
• 1:38:03
  This is the fifties.
• 1:38:04
  Yeah.
• 1:38:06
  It took us 80 years to get, to get to the Vera Rubin.
• 1:38:09
  So, yeah,
• 1:38:11
  So that's the first story.
• 1:38:13
  The second story, which is study number two, this is published in Nature Scientific Reports, and this is correlations with the atomic age.
• 1:38:23
  Transients in the Palomar Observatory Sky Survey may be associated with nuclear testing and reports of unidentified anomalous phenomenon.
• 1:38:33
  This is, it's got UAP in the title.
• 1:38:35
  It's got UAP in the title.
• 1:38:36
  Bro.
• 1:38:37
  In a nature subjournal.
• 1:38:38
  In a nature subjournal.
• 1:38:39
  Springer Publication.
• 1:38:41
  Which is a significant thing in and of itself given the aversion of...
• 1:38:48
  to doing anything related to this subject for most peer-reviewed journals.
• 1:38:53
  Yeah, I mean, I was pretty surprised myself.
• 1:38:55
  Yeah, when you were like, because I brought it up, you're like, well, what journal is it?
• 1:38:57
  And I was like, oh, and you're like, wait, really?
• 1:38:59
  Wait, nature, yeah.
• 1:39:00
  Fine, I guess if it's Springer, I thought we could do it.
• 1:39:04
  We'll cover it if it's Springer.
• 1:39:05
  Yeah, we'll cover it.
• 1:39:06
  So here what they're doing is they're doing more of a time-
• 1:39:10
  Sensitive approach, okay?
• 1:39:12
  They're looking at the rate of the transients that appear in those photographic plates that I talked about.
• 1:39:18
  And they're trying to correlate it with dates of above-ground nuclear weapons tests and daily counts of UAP reports, okay?
• 1:39:28
  So the timeline is, again, 1949 to 1957.
• 1:39:30
  Okay.
• 1:39:33
  They get a bunch of transient count.
• 1:39:36
  So for every single day, they've got photographic plates, and they count how many transients were found on each day.
• 1:39:43
  Each day also gets a binary flag on whether it's within a day of a nuclear test or not.
• 1:39:49
  Right?
• 1:39:50
  And back then, I think only three countries were doing nuclear tests, the Soviets, the U.S., and Great Britain.
• 1:39:56
  And then there's also daily UAP reports from the UFO CAT database.
• 1:40:02
  And these three things are what they're trying to relate.
• 1:40:07
  They use something called the generalized linear model.
• 1:40:09
  I'm very familiar with something called the generalized linear model because my PhD had a lot of that.
• 1:40:15
  Basically, these kinds of models are what you want to do when things are not just simply linearly correlated to one another.
• 1:40:24
  In my PhD specifically, I was trying to...
• 1:40:28
  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.
• 1:40:37
  All of these things can have varied responses that are not necessarily like, oh, if I'm looking this way, I just have more.
• 1:40:44
  Maybe it's like a weird response, right?
• 1:40:46
  Where only if he's looking this way, there's going to be more.
• 1:40:49
  So all of these different multimodal effects can be captured by a generalized linear model.
• 1:40:55
  And that's what they did.
• 1:40:57
  It can also simulate weird distributions that aren't just uniformly distributed or Gaussian distributed.
• 1:41:03
  If you've got a binomial distribution, which is like coin flips.
• 1:41:07
  Yep.
• 1:41:08
  If you've got, you know, did I see something or did I not?
• 1:41:11
  Things like that.
• 1:41:13
  It's very good at modeling those kinds of things, right?
• 1:41:15
  So there was a correlation to nuclear testing.
• 1:41:19
  So the transience occurrence in these photographic plates was 45% more likely within a three-day nuclear test window.
• 1:41:27
  Okay?
• 1:41:27
  And the main thing was it was 68% higher likelihood on a day after the test.
• 1:41:33
  Right?
• 1:41:34
  It's pretty significant.
• 1:41:35
  It's pretty significant.
• 1:41:36
  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.
• 1:41:49
  I don't really care for that as much.
• 1:41:51
  Like UAP reports to me are like...
• 1:41:54
  I don't know.
• 1:41:54
  Humans are fooled very easily.
• 1:41:56
  And UAP reports are just humans being like, I saw something that was crazy.
• 1:42:02
  I can't think of whether there are tests that I could do to see if this would work.
• 1:42:09
  For example, what if you assumed that on average, half of the UAP reports were nonsense?
• 1:42:16
  If you took out half,
• 1:42:17
  would the correlation still be there?
• 1:42:20
  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?
• 1:42:29
  Yeah.
• 1:42:30
  That's something that we used to do in my neuroscience lab in PhD.
• 1:42:34
  It's like...
• 1:42:35
  How sure are we that this variable is correlated to this variable?
• 1:42:39
  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?
• 1:42:47
  Yep.
• 1:42:48
  And given the shape of that plot, I'm not sure if I believe it.
• 1:42:51
  But the nuclear testing thing is just like, that's just there, right?
• 1:42:57
  Yeah, yeah.
• 1:42:59
  So those are the two reports.
• 1:43:02
  There's one report in PASP.
• 1:43:04
  That's physical evidence.
• 1:43:06
  You've got the Earth's shadow test that a substantial fraction of the transients are...
• 1:43:12
  sunlit objects, right?
• 1:43:13
  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?
• 1:43:26
  There are people who obviously...
• 1:43:31
  are skeptics, right?
• 1:43:33
  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.
• 1:43:46
  Yeah.
• 1:43:47
  Like, as of now, I haven't seen any follow-up scientific papers.
• 1:43:51
  Right.
• 1:43:52
  Because all of this data is publicly available.
• 1:43:56
  And I actually saw on the...
• 1:43:58
  On the two papers, it's readily available for people to look at.
• 1:44:03
  So I encourage anyone to do their own analysis, actually.
• 1:44:07
  I want to make a funny note.
• 1:44:09
  On the paper that I keep bringing up, within two weeks, because it was, was it?
• 1:44:13
  It was a week, dude.
• 1:44:14
  It was a week.
• 1:44:14
  Yeah.
• 1:44:14
  Within a week, because it was out of Cambridge or Oxford, one of the two.
• 1:44:17
  Yeah, Cambridge had the original paper, and then the guy at Oxford was like, nah.
• 1:44:20
  And within a week, Oxford came out with the rebuttal.
• 1:44:22
  Yeah.
• 1:44:23
  And then multiple came out.
• 1:44:24
  It's been more than a week.
• 1:44:26
  Yeah.
• 1:44:26
  Yeah.
• 1:44:27
  This paper is getting quite a bit of just, obviously, a lot of the science media is doing a variety of different coverage.
• 1:44:35
  But there's not yet necessarily been follow-ups as quickly.
• 1:44:39
  There's a number of reasons that could be true.
• 1:44:40
  Yeah, I mean, one of them, I think the big one is probably it's not an active scientific field.
• 1:44:46
  Like exoplanets and exoplanet researches, everyone's trying to make a name for themselves, right?
• 1:44:52
  They've got a career to look after.
• 1:44:54
  This one is not like...
• 1:44:55
  Something that, you know, okay, so you proved this wrong.
• 1:44:58
  It's not going to, like, give you a leg up in an astrophysical career.
• 1:45:01
  There's no incentive for people who are looking to climb a ladder to do a follow-up on it.
• 1:45:05
  Yeah, yeah, yeah.
• 1:45:07
  So that could be one of the reasons.
• 1:45:10
  But in any case, the Scientific American actually had an article about it that I was reading, and they showed some rebuttals, right?
• 1:45:19
  So there's the Hambly argument, which is from the University of Edinburgh.
• 1:45:23
  He's an expert, Nigel Hambly.
• 1:45:26
  He's an expert in photographic plate analysis.
• 1:45:29
  And...
• 1:45:30
  His main contention was that full width at half max, right?
• 1:45:32
  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.
• 1:45:41
  And there's skepticism that, you know, this is a classic signature of an emulsion or a flaw or a plate defect.
• 1:45:50
  And it's not a celestial object because a celestial object would have this kind of...
• 1:45:54
  noise.
• 1:45:56
  And one of the crucial things that he says is the ultimate arbiter is actually just the physical evidence.
• 1:46:01
  The plates are still around.
• 1:46:03
  We can take those original glass plates and examine them under a microscope and not examine the digital scans.
• 1:46:10
  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.
• 1:46:22
  So I think that's, you know, he's proposing a way to... To make it go away.
• 1:46:27
  To make it go away.
• 1:46:29
  I think that's good.
• 1:46:30
  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.
• 1:46:39
  Yeah, let us know.
• 1:46:40
  Let us know.
• 1:46:42
  And we already talked just previously about, even in the paper itself...
• 1:46:48
  a means by which that is not necessarily the likely outcome.
• 1:46:53
  However, you can remove it from the table.
• 1:46:55
  The earth shadow thing.
• 1:46:55
  Yes, the earth shadow point.
• 1:46:57
  You can still remove it entirely from being the cudgel that you're seeing being put out.
• 1:47:02
  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.
• 1:47:12
  Yeah, yeah, yeah.
• 1:47:13
  Which is kind of the issue.
• 1:47:14
  The earth shadow thing is pretty...
• 1:47:17
  Pretty substantive, I think.
• 1:47:18
  And so, however, the beauty of us still having the original plates is that it's falsifiable.
• 1:47:27
  Yes, exactly.
• 1:47:29
  There's another one that's proposed by Sean Kirkpatrick.
• 1:47:33
  Ooh, Dr. Sean Kirkpatrick.
• 1:47:37
  Former Pentagon AARO head.
• 1:47:40
  We like to call it ARO.
• 1:47:41
  Yeah.
• 1:47:41
  Arrow.
• 1:47:42
  Okay, Arrowhead.
• 1:47:45
  I guess that's like an anomaly office or something?
• 1:47:48
  That's the all-domain anomaly resolution office.
• 1:47:50
  Oh, okay, got it.
• 1:47:50
  That was created in 2022 as a result of Congress pushing on the executive branch to give them more information about... Again, because...
• 1:47:59
  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.
• 1:48:08
  I see.
• 1:48:09
  That was the impetus for why the office was established.
• 1:48:12
  And its mandate was twofold, which was unfortunate because –
• 1:48:17
  it was basically a mandate that had conflating outcomes.
• 1:48:21
  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.
• 1:48:34
  And if it's conventional, send it to the office that exists to deal with it.
• 1:48:38
  And if it's unconventional, figure out what we need to do to figure out what it is.
• 1:48:43
  That was one of its mandates.
• 1:48:44
  Yeah.
• 1:48:45
  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.
• 1:48:58
  The problem with that is you're asking the Department of Defense, which is not a public communications office, to do public communications.
• 1:49:05
  There's a reason why NASA – Yeah.
• 1:49:08
  Is its own agency.
• 1:49:09
  And you don't have NASA running out of the DOD because we would never get anything out of that structure.
• 1:49:15
  So there's sort of a conflate.
• 1:49:17
  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.
• 1:49:29
  We've basically seen none of it.
• 1:49:30
  Yeah, okay.
• 1:49:32
  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.
• 1:49:45
  Now, when you have these high-altitude nuclear detonations, you know, you can have radiation, particulate.
• 1:49:52
  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,
• 1:50:08
  reflect light in the way that we're seeing in these Palomar sky surveys.
• 1:50:12
  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,
• 1:50:24
  And one of the – there was a Q&A section where one of the people actually asked about the Chernkov radiation as a solution to the problem set.
• 1:50:33
  Someone can clip this and put the response.
• 1:50:36
  I don't think it's public yet.
• 1:50:37
  I didn't write it down, but she did have a response to that issue.
• 1:50:42
  Yeah.
• 1:50:43
  Which I unfortunately can't replicate in this context.
• 1:50:46
  But that was brought up by someone in the audience as the obvious next step in the progression.
• 1:50:52
  Yeah.
• 1:50:52
  Yeah.
• 1:50:53
  As an answer to it.
• 1:50:54
  Yeah, I can imagine, like, I mean, for Cherkhov radiation, it would probably be, like, streaky rather than single point sources.
• 1:51:03
  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.
• 1:51:12
  So, to me, the metallic junk is a better explanation than Cherkhov radiation or ionization, because those have very characteristic...
• 1:51:22
  signatures on photographic plates.
• 1:51:24
  But one of the challenges that Sean Kirkpatrick said was, you know, we've got GPS satellites now.
• 1:51:30
  Why don't we try to reproduce these results with the GPS satellites?
• 1:51:37
  I think it would...
• 1:51:39
  I think it's harder said than done because GPS, there's so many satellites now.
• 1:51:43
  Yeah, yeah, yeah.
• 1:51:44
  Right?
• 1:51:44
  It's not just like you're going to have so many streaks.
• 1:51:46
  Yeah, yeah.
• 1:51:48
  So for a 50-minute exposure, you're just going to get completely wrecked by Starlink.
• 1:51:53
  Right.
• 1:51:53
  There's another proposition by Elliot Gillum at the SETI Institute, Search for Extraterrestrial Intelligence.
• 1:52:02
  He said, you know, meteors traveling directly towards the telescope
• 1:52:06
  Could appear stationary dots, not streaks.
• 1:52:09
  Um, it's viable, but like now you got, it's traveling directly at the telescope.
• 1:52:14
  Like that, that, that volume level and simultaneous.
• 1:52:17
  Yeah.
• 1:52:17
  Yeah.
• 1:52:18
  Yeah.
• 1:52:18
  It's like simultaneous, multiple aligned.
• 1:52:20
  It's a little challenging.
• 1:52:22
  Yeah.
• 1:52:22
  It's a little challenging given the numbers.
• 1:52:24
  Um, but,
• 1:52:26
  There's a Princeton astrophysicist, Robert Lupton, who has this look elsewhere effect.
• 1:52:30
  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.
• 1:52:41
  And you can't actually use the Edmonds and George formula to do that kind of statistical significance.
• 1:52:49
  to figure out whether stuff isn't aligned because there's so many things that are in a line.
• 1:52:53
  You can easily fool yourself into thinking that the thing that isn't aligned is there for some reason, right?
• 1:53:01
  And the probabilities need to be corrected for the vast number of stuff that's out there.
• 1:53:06
  This goes back to your not looking at the alignment as the interesting part.
• 1:53:09
  Yeah, yeah, yeah.
• 1:53:10
  And I think there's a, I need a little bit more convincing on the alignment being something that's interesting.
• 1:53:17
  Okay.
• 1:53:17
  Right.
• 1:53:19
  But I mean, the debate is there, you know, 7.6 sigma deficit of transience in the Umbra.
• 1:53:27
  That's not something that I think we can ignore.
• 1:53:31
  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?
• 1:53:42
  Okay.
• 1:53:44
  It's certainly like giving me pause.
• 1:53:48
  Yeah.
• 1:53:49
  I'll be honest.
• 1:53:49
  Which I want to say for the audience, I've been working this guy for a long time.
• 1:53:54
  So to be giving him pause is statistically significant.
• 1:54:00
  Yeah, yeah, yeah, yeah, yeah.
• 1:54:02
  You probably have, like the number of times you've tried, I'm like, nah.
• 1:54:06
  I've tried so hard.
• 1:54:08
  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
• 1:54:16
  In a way that's compelling.
• 1:54:17
  Yeah.
• 1:54:19
  I mean, clearly, like, the team...
• 1:54:23
  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.
• 1:54:38
  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.
• 1:54:49
  Yeah.
• 1:54:49
  That's just like good math.
• 1:54:51
  Um,
• 1:54:52
  There is one that was kind of interesting.
• 1:54:55
  So Candidate 5, the five transients systematically in alignment, that happened on July 22nd, 1952.
• 1:55:01
  27th.
• 1:55:01
  Oh, yeah.
• 1:55:01
  July 27th, 1952.
• 1:55:03
  And they mentioned that this coincides with the...
• 1:55:12
  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.
• 1:55:27
  And apparently there were visual sightings, fighter jets were scrambled, there was a press conference from the Pentagon.
• 1:55:35
  So I didn't even know that this thing happened.
• 1:55:38
  So this is one of the weird, bro.
• 1:55:40
  This is one of the things that's so interesting.
• 1:55:41
  Why would they do that?
• 1:55:46
  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.
• 1:55:57
  That's very dense.
• 1:55:58
  And there's a lot of information there.
• 1:56:00
  And, you know, unless you have an entry point,
• 1:56:04
  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.
• 1:56:15
  And it is interesting to see how it cyclically comes back up.
• 1:56:21
  But that was Candidate 5.
• 1:56:23
  Yeah, and that was the one that was super aligned.
• 1:56:26
  I mean, I still think it's probably coincidence, but it's an interesting coincidence.
• 1:56:32
  I'll say that.
• 1:56:34
  The other thing they do at the end of the PASP paper...
• 1:56:40
  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.
• 1:57:00
  So then you get a really sharp image spec in your photographic plate.
• 1:57:08
  It's speculative at best.
• 1:57:13
  I think it's certainly interesting.
• 1:57:15
  I think the Earth's shadow thing is the one that is really sticking with me.
• 1:57:20
  That calculation is quite nice.
• 1:57:22
  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?
• 1:57:33
  Yeah, that's...
• 1:57:35
  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.
• 1:57:42
  So, yeah, I thought it was pretty cool.
• 1:57:44
  I think next steps would be to just look at the photographic plate.
• 1:57:48
  Make sure that it's not a defect.
• 1:57:49
  Take a look at the plate, not a defect.
• 1:57:52
  And then from there, if proven to be the case, now we can have the next stage of the conversation.
• 1:57:57
  Yeah, which is like, you know, with modern surveys, perhaps there's a way to get...
• 1:58:02
  An idea of how often this is happening in Verirubin.
• 1:58:04
  Verirubin's completely public.
• 1:58:09
  All of the data is public immediately after it's taken.
• 1:58:12
  And it's up and running right now.
• 1:58:13
  Yeah, yeah.
• 1:58:14
  So we have already three months of data from the Verirubin.
• 1:58:17
  If you've not already listened, take a look back to our previous episode where we covered the Verirubin Observatory.
• 1:58:24
  Why it matters.
• 1:58:25
  It's another sky survey, but it is...
• 1:58:27
  Just light years in terms of the structure and substance of how it works and what we talked about, the immediate data availability.
• 1:58:36
  No embargo.
• 1:58:38
  Again, some people will be like, well, you know, government's going to try to wipe away their classified projects that are –
• 1:58:48
  Let's say three months.
• 1:58:50
  Yeah.
• 1:58:50
  Three months meaning 100 days, meaning there's 30 photos of the entire night sky already.
• 1:58:55
  Right.
• 1:58:55
  Right.
• 1:58:56
  Palomar took years for one photo of the entire night sky.
• 1:58:59
  This already has 30.
• 1:59:01
  It's an incredible, the largest...
• 1:59:04
  Camera?
• 1:59:05
  Yeah.
• 1:59:05
  The largest camera in the world.
• 1:59:06
  In the world.
• 1:59:09
  Creating the biggest amount of data.
• 1:59:11
  Yeah.
• 1:59:12
  And one of the most incredible data processing and distribution platforms.
• 1:59:16
  Yeah.
• 1:59:16
  It's insane.
• 1:59:17
  It's just incredible.
• 1:59:19
  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.
• 1:59:31
  and learnings to it in addition to continuing to validate and replicate using the plates from the Palomar Sky Survey as well.
• 1:59:42
  Like there's multiple avenues of exploration here around this idea of these transient light artifacts that are occurring.
• 1:59:53
  I really was excited to just get your perspective on this.
• 1:59:58
  Yeah.
• 1:59:58
  I thought it was a cool set of papers, yeah.
• 2:00:02
  Definitely the PISP one I liked.
• 2:00:03
  And just also talking about it from first principles.
• 2:00:06
  Again, yeah, I think it's important to understand because this is historical data, we're not used to analyzing that, right?
• 2:00:15
  And the analog...
• 2:00:17
  Nature of it.
• 2:00:19
  It's not digital, right?
• 2:00:20
  You're not sticking a CCD in the back of a telescope.
• 2:00:22
  This is, at the end of the day, a chemical reaction that's giving us something.
• 2:00:26
  So it's important to understand the whole chain of custody from photon to pixel.
• 2:00:32
  Right, right.
• 2:00:33
  And how do we even get to pixel from photon?
• 2:00:34
  Yeah.
• 2:00:36
  an incredible set of stories this week.
• 2:00:39
  We did our lucky number three.
• 2:00:41
  We started with our generative AI for cancer detection out of university of Toronto.
• 2:00:47
  Again, sorry again about the loss, but you're going to save the lives.
• 2:00:50
  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.
• 2:01:01
  We're making blind people see again.
• 2:01:03
  Yeah.
• 2:01:03
  Yeah.
• 2:01:04
  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
• 2:01:31
  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
• 2:02:00
  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
• 2:02:29
  from the sun and now it's on the other side of the sun.
• 2:02:31
  So we can actually see it again from the earth.
• 2:02:34
  It's on its way to Jupiter, but we have no data that we can talk about.
• 2:02:39
  So until we get the data, we won't cover it.
• 2:02:41
  Yeah.
• 2:02:42
  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,
• 2:02:54
  Or not.
• 2:02:55
  Thingy or not until we get the data.
• 2:02:58
  Yeah.
• 2:02:59
  So we hear you guys.
• 2:03:01
  You want us to cover 3i Atlas.
• 2:03:03
  Open up the government.
• 2:03:04
  Give us the data and we will talk about it.
• 2:03:06
  Yeah.
• 2:03:07
  I am your host, Lester Nari, joined as always by my co-host and our resident PhD, Krishna Chowdhury.
• 2:03:14
  This is From First Principles.
• 2:03:16
  We'll see y'all next week.