Latest / From First Principles / 5,000-Year-Old Bacteria, Solar Storms, Dogs, and Meta’s AI War (EP. 32)

Transcript

5,000-Year-Old Bacteria, Solar Storms, Dogs, and Meta’s AI War (EP. 32) · From First Principles

• 0:00
  Frozen for 5,000 years, this ice cave bacterium resists modern antibiotics, which sounds a little bit scary.
• 0:08
  It's very speculative, but even the fact that the sun could affect earthquakes is a very interesting... That is, Jan LeCun, who has recently left Facebook or Meta in early 2026, and has now raised a billion dollars to build AI that understands the physical world.
• 0:29
  world.
• 0:29
  Hello, Internet.
• 0:30
  This is your captain speaking, Lester Nare, joined as always by my co-host and our resident PhD, Krishna Chowdhury.
• 0:38
  We are in our first standalone rundown episode.
• 0:41
  We're going to cover several stories just at a high level, have a fun conversation.
• 0:46
  These are stories we didn't get to to do a deep dive this week, but we still want to talk about it.
• 0:51
  And get you guys into your weekend with some fun, exciting talking points with a special segment in the middle, which is going to be a surprise for our resident PhD.
• 1:02
  Today we're going to cover microbiology on bacteria that was found in frozen caves 5,000 years ago, solar storms and earthquakes and how they may be related, and how they may be related.
• 1:13
  Some of the shared genetic roots between us and dogs, and we will wrap up with an AI story about how the battle for control at meta AI has now led to the start of a new billion dollar AI company coming at a different angle than large language models as we know it.
• 1:32
  We will be learning about the science not quite from the ground up, but at least from the top of the surface, because this is From First Principles.
• 1:55
  Our first story for the rundown today is out of the Institute of Biology in Bucharest of the Romanian Academy, and it was in the frontiers in microbiology.
• 2:05
  Frozen for 5,000 years, this ice cave bacterium resists modern antibiotics, which sounds a little bit scary.
• 2:13
  Yeah, dude, this one was pretty weird.
• 2:17
  There is a cave in Romania.
• 2:20
  It is called the Scari Soara Cave.
• 2:23
  Okay.
• 2:24
  It's got this giant glacier that's about the size of 40 Olympic swimming pools.
• 2:29
  And that glacier began to form about 13,000 years ago.
• 2:33
  And what you can do is you can drill cores into that ice and then extract those samples.
• 2:39
  And there's bacteria that are preserved in that ice.
• 2:43
  And bacteria are notorious for surviving being frozen for thousands of years.
• 2:48
  So we can bring back to life bacteria that have been frozen for thousands of years.
• 2:53
  This particular sample is from the 5,000-year-old layer of that glacier.
• 3:00
  They took it out, they revived it, and they tested it on modern antibiotics.
• 3:07
  They tested 28, and this bacteria was resistant to 10.
• 3:11
  Not great.
• 3:12
  Yeah.
• 3:13
  It's really quite strange on the face of it.
• 3:16
  Okay.
• 3:16
  Because modern antibiotics are, by definition, modern, right?
• 3:21
  So we've developed these...
• 3:25
  you know, antibacterial medication such that it kills bacteria.
• 3:30
  And these bacteria that are 5,000 years old still have a kind of defense to stuff that we made today.
• 3:36
  This is in line with a lot of the antibacterial resistance that we're faced with as humanity, right?
• 3:45
  We've been using antibiotics for hundreds of years now.
• 3:48
  The first real big one was penicillin by Alexander Fleming before World War I.
• 3:54
  Because we've been going at it, bacteria has evolved to fight against these antibiotics, right?
• 4:00
  They've figured out ways to get around it.
• 4:03
  And the fact that 10 of these 28 did not work on bacteria that were 5,000 years old suggests that these bacteria were already resistant in the first place, right?
• 4:15
  Okay, fine, 18 worked.
• 4:16
  But still, the 10 is kind of weird.
• 4:18
  Yeah, yeah, yeah.
• 4:19
  It kind of dovetails with the story we just talked about.
• 4:24
  As it relates to bringing back samples from space.
• 4:26
  Yeah.
• 4:27
  And the concern that arises because we don't know what will happen.
• 4:32
  Yeah.
• 4:32
  If there happens to be, again, that's a little bit, this is real world.
• 4:35
  We know that it's there.
• 4:36
  Yeah.
• 4:38
  We've tested it.
• 4:39
  And they're basically pre, they have pre-immunity.
• 4:42
  Yeah.
• 4:43
  They were never our current antibiotic infrastructure.
• 4:45
  They were never exposed to the current antibiotics.
• 4:47
  Right.
• 4:48
  And yet they're still already immune.
• 4:50
  Right.
• 4:50
  Right.
• 4:50
  It's quite weird.
• 4:51
  Right.
• 4:52
  I don't think we should be too worried about these glacial bacteria because these are called psychobactors.
• 4:59
  Psychobacteria, which means they're cold-loving.
• 5:01
  They're not going to survive in the 98 degrees Fahrenheit environment of the body, right?
• 5:10
  But it's still quite a nice little problem in molecular biology.
• 5:18
  How could this even happen?
• 5:20
  Right.
• 5:20
  And...
• 5:21
  When you look at how bacteria actually evolve, they evolve together.
• 5:25
  Bacteria are incredibly dynamic with their genomes.
• 5:28
  They swap genes from one to another all the time.
• 5:32
  That's actually how we transform bacteria.
• 5:34
  A lot of times in an undergrad lab, if you want to make a bacteria glow in the dark or something with a GFP, which is this fluorescent molecule, what you can do is put in a plasmid, which is a circular piece of DNA that has the GFP code.
• 5:48
  Mm-hmm.
• 5:49
  You put that in, and then you heat shock the bacteria.
• 5:52
  There's something that we've talked about a lot, where you just take the bacteria and you put it in hot water.
• 5:57
  And then the bacteria kind of freak out, and they're like, what are we doing?
• 6:00
  They ingest all sorts of DNA from their environment, and they try to transcribe that and see if that'll work.
• 6:06
  So bacteria are swapping DNA between themselves and the environment all the time, and perhaps through the swapping, you get this kind of drug resistance.
• 6:14
  What's really cool about this process
• 6:16
  story is that it actually the existence of this reveals how antibiotic resistance evolves naturally so that's one thing the other thing is that this ancient strain also showed the inability the so this ancient strain showed the ability to inhibit the growth of other modern superbugs right
• 6:36
  So kind of like in penicillin, when Alexander Fleming discovered penicillin, it was because it was a natural antibiotic that he had seen there was a mold that was growing in his Petri dish.
• 6:46
  And around that mold, there was no bacteria.
• 6:48
  And he was like, the mold is creating something that is killing the bacteria.
• 6:52
  Well, in this case, this ancient strain...
• 6:55
  is able to create compounds that will kill modern superbugs, right?
• 7:01
  So it possesses some kind of enzymatic gene that can hold existing biochemical potential, right?
• 7:09
  Existing pharmaceutical potential.
• 7:11
  And so it tells us that maybe we need to look at these kinds of natural places, at isolated populations of bacteria to figure out how to beat modern-day superbugs that are now resistant to our bacteria.
• 7:25
  antibacterial compounds, antibiotic compounds.
• 7:28
  But maybe these guys have something that we can use to defeat them.
• 7:33
  You know, it's like that Thor, I can't defeat you, but she can thing.
• 7:39
  But the she can thing is like these 5,000-year-old bacteria.
• 7:43
  Right.
• 7:43
  We can look to the past –
• 7:49
  to help inform us about how to fight in the present and in the future in a way that was maybe somewhat unexpected in that these bacteria that are 5,000 years old contain sort of an active ingredient that is relevant to sort of a pharmacological, pharmaceutical application today.
• 8:09
  And historically, most antibiotics are developed from bacteria and fungi
• 8:13
  that live in the soil.
• 8:14
  I see.
• 8:15
  And, like, they produce natural compounds because over there, down there, it's like D-Day every day, right?
• 8:21
  It's just World War all the time.
• 8:22
  Everyone is trying to eat everyone.
• 8:24
  And in recent times, those pathogens that we found from existing soil bacteria no longer work.
• 8:29
  Well, we can, you know, go back in time.
• 8:32
  Yeah.
• 8:32
  I think it's a pretty cool story at the heart of it.
• 8:36
  And it gives a new meaning to the phrase natural immunity.
• 8:41
  Yeah.
• 8:41
  And...
• 8:43
  If we can scale up the same process that allows you to freeze bacteria and reanimate them 5,000 years later, maybe Walt Disney has a chance.
• 8:51
  Because if I recall correctly, he's currently cryofrozen somewhere deep underneath Burbank's headquarters.
• 8:59
  That is rumor.
• 9:00
  That's rumor, of course.
• 9:01
  I don't know.
• 9:01
  We don't know anything about it.
• 9:03
  So not that to be true.
• 9:04
  But this is, again, a new research paper out of the Institute of Biology in Bucharest of the Romanian Academy.
• 9:12
  It was in Frontiers in Microbiology.
• 9:15
  Fun little microbiology story to get us started on our rundown.
• 9:20
  Number two.
• 9:22
  on our rundown today is can solar storms trigger earthquakes?
• 9:27
  Yes.
• 9:27
  And scientists propose a surprising link.
• 9:30
  This was actually covered by Science Daily.
• 9:32
  It was out of Kyoto University.
• 9:36
  And there's a huge solar storm monitoring community on a variety of social media channels.
• 9:43
  Every time there is, I guess, a coronal mass ejection or similar, there's this, like, ranking system that's at the Richter scale,
• 9:53
  Over 9,000 hunkered down prepper.
• 9:56
  But there's a potential link here to actually geological activity.
• 10:01
  Yeah, and that one I was pretty surprised that there is even a theoretical argument that you can make here.
• 10:07
  And this is mostly a theoretical argument.
• 10:08
  This is a theoretical paper.
• 10:11
  What they're doing is showing a surprising theoretical connection between space weather and earthquakes, geological activity on Earth.
• 10:19
  Now, how does that even...
• 10:21
  What is the mechanism there, right, is what I always wonder.
• 10:24
  So here's the idea.
• 10:26
  Faults on the Earth are poised at a kind of critical spot where there's a lot of stress on them and they're ready to snap, okay?
• 10:36
  Because they've been pushed to their brink and, like...
• 10:40
  The faults on the Earth are these, you know, plates on the Earth's surface that are contiguous blocks of the Earth's crust that are moving around because magma underneath is, like, flowing in a liquid motion.
• 10:53
  And so, you know, you move around just like how ice on a frozen lake sort of moves around based on currents underneath.
• 11:00
  So...
• 11:01
  Why would that be affected by solar activity is the idea.
• 11:06
  Well, solar activity is effectively a bunch of charged particles coming at us.
• 11:11
  That's effectively what it is, right?
• 11:13
  There's a bunch of electrons, let's say, that come at us, and those electrons are going to create a massive change in the ionosphere.
• 11:22
  The ionosphere being the outside ions around the Earth's atmosphere that are built by the magnetic field that sort of
• 11:32
  you know, shield us from space weather.
• 11:36
  Now that ionosphere is going to get changed up a lot because there's going to be a huge influx of charged particles from the sun.
• 11:43
  That's going to change the electric field on the earth.
• 11:45
  The earth's crust on the other hand has all of these plates that are sort of at this critical point.
• 11:51
  And in between all these plates inside the earth's crust, there's a lot of water that is in a super critical state because it's under high pressure.
• 11:58
  It's under high stress.
• 11:59
  Um,
• 12:00
  And that water has a bunch of ions.
• 12:03
  So the electric field on the outside is going to move around the electric field inside these cracks in the Earth's crust.
• 12:09
  And that could potentially create enough of a force to actually snap and create high earthquakes, you know, high damage earthquakes.
• 12:20
  The idea is that there is...
• 12:23
  These fields that are all around us all the time and interacting have impacts in the physical nature in which we see and we interact, the tangible things we can touch and feel.
• 12:37
  And at the scale of the sun and the scale of these solar storms, the amount of the change in fields...
• 12:44
  in the ionosphere can potentially have a subsequent impact on plate tectonics.
• 12:50
  Yeah.
• 12:50
  And, and how they move, which again, big enough solar storm means it would not only trigger things like frying electronics and all that other stuff.
• 12:57
  Yeah.
• 12:58
  But could actually physically impact the earth, the earth.
• 13:02
  Yeah, and there's like weak correlation between solar activity and geological activity.
• 13:10
  If you go to 2.5, there's a weak correlation that was out in Nature Scientific Reports in 2020.
• 13:18
  Again, it's not that big, and this paper is not claiming, this current paper out in 2026 is not claiming that, you know, solar flares directly cause all quakes.
• 13:27
  But the research provides this, like, kind of new mechanism to explain maybe unusual ionospheric activity and that connection with major quakes.
• 13:36
  Because in 2024, there was the Nodo Peninsula quake.
• 13:40
  And that was preceded by a really large solar storm.
• 13:44
  And so there's these coincident sort of, we don't have enough data to like, you know, the sigma isn't down to five.
• 13:50
  Yes.
• 13:50
  But...
• 13:52
  It's there, and this is exploring some of the mechanisms.
• 13:55
  I think that was pretty cool.
• 13:57
  It's very speculative, but even the fact that the sun could affect earthquakes is very interesting.
• 14:04
  That is, and I think the plot of a future upcoming Jerry Bruckheimer produced disaster movie.
• 14:11
  Jerry, if you want a scientific advisor, I think I know someone I can suggest for that.
• 14:15
  I've done it before, so...
• 14:20
  Great.
• 14:20
  So this is, again, out of Kyoto University, which is fine.
• 14:22
  I mean, not funny.
• 14:23
  It's interesting.
• 14:23
  Obviously, Japan has had very dramatic earthquakes.
• 14:26
  Yeah.
• 14:27
  If we think of Fukushima, etc.
• 14:28
  Yeah.
• 14:29
  So they have a very concerted effort to try to figure it out.
• 14:32
  And this was in the International Journal of Plasma, Environmental Science and Technology.
• 14:36
  Yes.
• 14:37
  Yes.
• 14:37
  Makes sense.
• 14:38
  Makes sense.
• 14:39
  This is going to be fun.
• 14:41
  All right, here we go.
• 14:42
  And we're going to try this.
• 14:43
  And if it doesn't work, you're not going to hear this on the final episode.
• 14:49
  But we are going to start a new game show segment called Are You Smarter Than a Scientist?
• 14:58
  Welcome, ladies and gentlemen, our resident PhD, Krishna Chowdhury, who will be our scientist of the day.
• 15:05
  And we will have one question, family feud style, for you to answer.
• 15:09
  And today's segment question for Are You Smarter Than a Scientist is, can you name the 10 most abundant elements in the human body?
• 15:20
  We have 10 on this list.
• 15:21
  Okay.
• 15:22
  Okay.
• 15:23
  All right.
• 15:23
  Okay.
• 15:24
  So I'm not sure if when you mean abundant, if you mean by number or by mass.
• 15:30
  So let's test it out and start with hydrogen.
• 15:34
  Okay.
• 15:35
  So we have you on the board with hydrogen.
• 15:39
  Okay.
• 15:39
  So hydrogen is number three, which means that you are going for by mass and not by number.
• 15:46
  Maybe.
• 15:47
  So let's go with oxygen.
• 15:48
  That should be number one.
• 15:50
  Is it number one, ladies and gentlemen?
• 15:53
  And oxygen is number one.
• 15:55
  Okay, so we got H2O.
• 15:56
  70% of the body is H2O, so that tracks.
• 16:00
  Let's go with carbon.
• 16:01
  Okay, carbon.
• 16:03
  Is carbon on the board?
• 16:05
  Yes.
• 16:06
  You got the top three.
• 16:07
  Okay.
• 16:07
  And then nitrogen.
• 16:09
  That should round up four.
• 16:10
  Okay.
• 16:10
  So you think that nitrogen is going to be number four?
• 16:13
  Yes.
• 16:14
  Is nitrogen number four?
• 16:15
  You are on a roll.
• 16:17
  We have our top four.
• 16:18
  Okay.
• 16:18
  So we got our top four.
• 16:19
  All right.
• 16:20
  That was the easy stuff.
• 16:21
  I expected you to get that.
• 16:22
  Yeah.
• 16:22
  So now let's do... Now, I don't know if this is going to be number five, but I know it should be phosphorus because nucleic acids, ATP...
• 16:31
  Okay, so phosphorus.
• 16:33
  The phospholipids of our cell membrane.
• 16:37
  Not number five, but still on the board at number six.
• 16:40
  Number six.
• 16:41
  So there's something in between nitrogen and phosphorus.
• 16:45
  Only thing I can think of is sulfur.
• 16:49
  Okay.
• 16:50
  Let's go with sulfur.
• 16:51
  Okay.
• 16:52
  That's in proteins.
• 16:54
  We may have sulfur on the board, but it's going to be at the number eight spot.
• 16:58
  Okay, what am I doing wrong?
• 17:02
  Um, wow.
• 17:02
  What is number five?
• 17:05
  Nitrogen.
• 17:06
  So, so right now we have oxygen.
• 17:08
  Number one, carbon two, hydrogen three, nitrogen four, phosphorus six and sulfur eight.
• 17:15
  We are missing five, seven, nine, and 10.
• 17:19
  No cheating.
• 17:20
  All right.
• 17:21
  Um,
• 17:23
  Let's do...
• 17:25
  So sulfur's already in there.
• 17:26
  Wow.
• 17:28
  Oh, calcium for Marbones.
• 17:31
  We have number five on the board with calcium.
• 17:35
  All right.
• 17:35
  There we go.
• 17:36
  All right.
• 17:37
  Now let's do...
• 17:38
  Reminder, you have three strikes.
• 17:40
  Oh, I do have three strikes.
• 17:41
  Okay.
• 17:42
  I don't want to use any, though.
• 17:43
  You know me.
• 17:45
  Okay.
• 17:46
  So I think now we should get into ions.
• 17:49
  Okay.
• 17:50
  So I'm going to go with the bread and butter of our nervous system and ions in general.
• 17:56
  So let's go with sodium first.
• 17:58
  That's why we like salt.
• 18:00
  Sodium is on the board at number nine.
• 18:03
  We have two left, number seven and number 10.
• 18:07
  Let's do potassium.
• 18:09
  Is potassium on the board?
• 18:12
  No.
• 18:14
  It is at number seven.
• 18:15
  So we got sodium and potassium.
• 18:17
  That's our NAK channels for the neuroscientists out there.
• 18:22
  One left, our number 10.
• 18:24
  This might be a tough one.
• 18:25
  This one's a tough one.
• 18:26
  This might be a tough one.
• 18:29
  Okay.
• 18:30
  So...
• 18:32
  um can be oxygen carbon hydrogen nitrogen calcium phosphorus potassium sulfur and sodium we are looking for the 10 most abundant elements in the human body if you at home have gotten this before krishna you may be smarter than a scientist yeah i don't know okay i'm gonna guess on this one okay you have to go with iron okay because of our blood hemoglobin iron is it on the board
• 19:01
  Iron is not.
• 19:02
  Okay.
• 19:03
  Shit.
• 19:04
  Not on the board.
• 19:04
  Strike number one.
• 19:08
  Got two more strikes in the chamber.
• 19:13
  We do not have phone a friend.
• 19:15
  We do not have any escape routes.
• 19:20
  We already got calcium.
• 19:22
  I put you on the spot today.
• 19:23
  I know.
• 19:23
  This one's a hard one.
• 19:25
  Let's go with zinc.
• 19:26
  Zinc is heavy.
• 19:27
  And I know that we need it.
• 19:30
  Okay.
• 19:30
  I just don't know for what.
• 19:31
  I'm going to go with zinc.
• 19:33
  Number 10.
• 19:34
  So even if we don't need a lot.
• 19:35
  Is it zinc?
• 19:36
  Zinc.
• 19:38
  No.
• 19:39
  Oh, my gosh.
• 19:40
  It is not, Zik.
• 19:41
  All right.
• 19:43
  One more strike left for the 10 most abundant elements in the human body.
• 19:50
  All right.
• 19:52
  So I'm thinking it's either iodine because I know we need iodine.
• 19:56
  That's why we need iodized salt or magnesium.
• 19:59
  I know a lot of people take supplements of magnesium, but people also take iron supplements, and that wasn't on it.
• 20:05
  Let's go with magnesium.
• 20:07
  Final answer?
• 20:08
  Yeah, let's do it.
• 20:09
  Final answer for the number 10 spot.
• 20:14
  10 most abundance elements in the body.
• 20:16
  Is it?
• 20:17
  No.
• 20:18
  All right, what is it?
• 20:20
  It is not our number 10 spot.
• 20:23
  If you're at home, plug in your guesses now.
• 20:27
  For our number 10 spot, the answer was chlorine.
• 20:32
  How did I...
• 20:35
  Oh, my God.
• 20:35
  I even said salt.
• 20:37
  Sodium chloride.
• 20:38
  Wow.
• 20:40
  So that's ridiculous.
• 20:41
  This was our first rendition of Are You Smarter Than a Scientist?
• 20:47
  If you got all 10, you may have one out on this week one.
• 20:53
  So this was our first go around doing this.
• 20:56
  I think it was pretty...
• 20:59
  But that's just my personal opinion.
• 21:03
  I now have also access to a soundboard, so this can get very outrageous very quickly.
• 21:08
  I will be sure to hold it in as much as possible.
• 21:13
  But it looks like, for the most part, everything went smoothly on the board itself.
• 21:18
  I think this could be super fun.
• 21:19
  That would be really fun.
• 21:20
  So again, let us know in the comments if you are intrigued by Are You Smarter Than Scientists and would like to see us bring that segment into the rundown episode on a regular basis.
• 21:33
  We are going to now go into our third story of the day, which was the one about the link, a shared genetic link between humans and dogs.
• 21:45
  And this is about golden retriever genes linked to anxiety, aggression, and intelligence in humans out of the University of Cambridge and the Morris Animal Foundation in the proceedings of the National Academy of Sciences.
• 22:00
  Yeah, this was a really cool one.
• 22:03
  They studied 1,300 golden retrievers, and it turns out that not all golden retrievers have the same personality.
• 22:09
  I mean, anyone who has multiple golden retrievers or has seen multiple golden retrievers knows this.
• 22:14
  They're not like a one-brand type dog.
• 22:16
  There's dogs that are shy.
• 22:18
  There are dogs that are, you know...
• 22:21
  a little timid.
• 22:22
  There's dogs that are very, very social.
• 22:24
  So there's enough genetic variation within golden retrievers where when these scientists studied 1,300 golden retrievers, they could figure out genes that were related to certain traits.
• 22:35
  Each of these 1,300 golden retrievers, the owners of those dogs got like a little questionnaire about like, you know, personality traits for your dog.
• 22:44
  And then they could correlate using the genetic data with,
• 22:47
  Certain genes with certain traits.
• 22:49
  And then when they went and looked at humans and those human traits, the same genes mapped to the similar human traits.
• 22:57
  Interesting.
• 22:58
  That's the idea.
• 22:59
  Interesting.
• 22:59
  So there's two kind of points of data.
• 23:01
  There's sort of a quantitative data, which is the genetic piece.
• 23:06
  And there's like a qualitative data point, which is the survey.
• 23:09
  Exactly.
• 23:10
  And there was the survey.
• 23:11
  There was a reference point both for dogs and personality traits on the qualitative side.
• 23:16
  Yeah.
• 23:16
  And humans.
• 23:17
  And when they looked at the genes that were being expressed that mapped to the personalities in dogs.
• 23:23
  Yeah.
• 23:24
  Based on the feedback from the owners.
• 23:26
  What's interesting is that if you point to those same genes in humans...
• 23:30
  The same qualitative data matches up.
• 23:33
  Yeah, and you can map the genes together because even though obviously the golden retriever gene and the human gene are not going to be identical, they're dogs, they're humans.
• 23:43
  But you can look at the sequence of nucleotides, the ATGC, ATGC, and you can find overlap where like this gene is like there's only very few mutations, right?
• 23:52
  So it's definitely the same gene.
• 23:53
  And then from that, they can figure out these genes that are actually related.
• 23:56
  There's one gene, for example, PTPN1.
• 24:01
  to aggression towards other dogs in golden retrievers, and it's associated with intelligence and depression in humans.
• 24:09
  So not same-same, but it's really kind of this meta-behavioral state or personality type that these genes are controlling.
• 24:18
  They were very careful about making that distinction, that this is not like a one-to-one mapping, but it's rather like...
• 24:26
  A shared biological route for overarching personality.
• 24:32
  It's more meant to be a framework versus like a purely like, for lack of a better term, like diagnostic.
• 24:38
  Exactly.
• 24:38
  A discrete diagnostic tool.
• 24:40
  Yeah.
• 24:40
  And it helps create a scaffolding by which to kind of architect the structures that.
• 24:47
  And then you can swap out the variables within those structures as makes sense.
• 24:50
  Yeah.
• 24:50
  And it also like leads to like insights and training dogs, for example, because like certain dogs might just be a certain way because of the genetic makeup.
• 25:01
  Right.
• 25:01
  And it has nothing to do with like their history or like so on and so forth.
• 25:06
  Here's a quote from one of the lead researchers.
• 25:08
  Yeah.
• 25:09
  If your golden retriever cowers behind the sofa every time the doorbell rings, perhaps you might have a bit more empathy if you know they're genetically driven to feel sensitive and anxious.
• 25:21
  Love that from this researcher from the University of Cambridge, Ana Morris Nuevo.
• 25:27
  It's funny.
• 25:28
  As a...
• 25:32
  someone who has three dogs who have very, very, very distinct and different personalities.
• 25:42
  It's an interesting just to think about.
• 25:44
  And also, as we continue to sort of in a variety, because we've touched a lot of genetic stories, especially in the last couple of weeks.
• 25:54
  And just the breadth and depth of the increase in our understanding of like this fundamental, like, you know, the instruction manual for how our bodies and all like living beings like create the stuff, the factories.
• 26:07
  Yeah.
• 26:07
  Right.
• 26:08
  And the IP that generates.
• 26:10
  It's as we continue to learn more, it.
• 26:15
  It's not only the understanding piece, which is the first step, but that understanding is also in already being applied because there's a whole variety of things we can now do because we actually know the map.
• 26:28
  Yep.
• 26:29
  We actually have that baseline.
• 26:31
  Reference point.
• 26:33
  And we always love a dog story.
• 26:34
  Yes, always.
• 26:35
  Whenever I see dog stories in science news, I'm like, yep, we're going to cover it.
• 26:39
  The corgi story continues to be, I think, our most popular dog story.
• 26:43
  Yes.
• 26:43
  And then the whale, the pods attacking the... That's our biggest...
• 26:48
  animal.
• 26:48
  It's actually our biggest story ever, which I still can't...
• 26:51
  It's still so surprising to me.
• 26:54
  We're going to wrap up the rundown today.
• 26:55
  Again, super casual Friday.
• 26:57
  Almost summer Fridays.
• 26:58
  Here it's 92 degrees in Los Angeles, California.
• 27:01
  That's right.
• 27:02
  Currently, which is...
• 27:04
  Uh, that's an average March.
• 27:06
  Friday.
• 27:07
  So, uh, we're finally over the rain, I think apparently, uh, which is great.
• 27:13
  And our last story is, is, you know, interesting because it dovetails with a lot of the AI stories we've talked about.
• 27:19
  And there's both a human, uh,
• 27:22
  personality side to the story and sort of like a an actual science part to the story and this is related to uh jan lacun who has recently left facebook or meta uh in early 2026 and has now raised a billion dollars to build ai that understands the physical world and he
• 27:46
  This is related to – this was covered by Wired, and there's a lot of background to this story.
• 27:51
  I'll kind of just quickly paint a picture before we kind of get into the physical world AI piece.
• 27:57
  In terms of – if you've listened to any of the Frontier model CEOs over the last 12 months, so this is Sardemis Asabis at Google DeepMind.
• 28:08
  who runs all of Google's AI products, Gemini, everything that's integrated with Google Workspace.
• 28:14
  You have Dario Amadei, CEO of Anthropic, who has the Claude and Claude Code and co-work.
• 28:22
  They've been very much in the zeitgeist and in the mainstream because of their ongoing kerfuffle with the Pentagon over usage of Claude within military applications, specifically autonomous weapons and mass surveillance equipment.
• 28:38
  primarily focused on civilian, but you could argue generally.
• 28:42
  And particularly with Davos, the Davos speeches that just happened, all of them were there.
• 28:48
  And talking about all the journalists were like, where is this going and when are we going to get to, quote, AGI, Artificial General Intelligence, or ASI, Artificial Superintelligence, and can simply scaling the current architectures, meaning putting more compute and more data into it,
• 29:06
  Get us there.
• 29:07
  Yeah.
• 29:08
  Or are we missing a fundamental architecture or concept that needs to be included in order to actually make that jump?
• 29:17
  And Jan LeCun has been very vocal about his perspective that the scaling laws are not going to get you to AGI.
• 29:28
  Yeah, he's been very vocal about that.
• 29:30
  And he's had some really nice little analogies about why he thinks that is.
• 29:34
  Yes.
• 29:35
  He equates it to the models of Ptolemy on the epicycles.
• 29:40
  Yes.
• 29:40
  And how before Copernicus, everyone was trying to fit the Earth as the center of the universe.
• 29:46
  And so they came up with...
• 29:48
  orbits around Earth, but obviously Mars does this weird retrograde motion where it goes backwards.
• 29:54
  And so Ptolemy decided, okay, I'm going to put an epicycle, which is Mars has a big circle around the Earth, but then around that circle, there's a little circle that Mars goes around.
• 30:03
  So every once in a while on that little circle, it's going to go back.
• 30:06
  But then as you get closer and closer and the data gets harder and harder, you just start adding more and more epicycles.
• 30:12
  And it
• 30:17
  For infinitely, the more epicycles you add, the closer and closer you'll get to the ground truth.
• 30:22
  And you'll never discover Kepler's ellipses if you don't have a fundamental change.
• 30:27
  And Janne Klun is saying that basically our scaling hypothesis, which is just make the parameters more and more, we're just adding epicycles.
• 30:36
  Right.
• 30:36
  fitting data but the underlying parameter of newtonian gravity and the one over r squared dependence we're missing that entirely and i like that little that's a great you know that's a great and now like reference point yeah and just to briefly kind of touch on so you know people talk about chat gbt and claw and all this and in the fundamental architecture that is
• 31:01
  And undergirding them is this idea of large language models.
• 31:05
  And so the idea is you're optimizing language as an abstract or as a construct.
• 31:12
  And both Demis Asabas and Jan LeCun in the last 6 to 12 months have discussed the importance of world models.
• 31:20
  Yeah.
• 31:21
  This idea that language models are not actually having this fundamental understanding of the physics of reality, which is what has maybe made it a little bit more difficult to get image generation and video generation that out of the box feels real.
• 31:39
  What seems to have happened is these LLMs have sort of created this like...
• 31:44
  abstraction of what they understand physics to be.
• 31:48
  But it is not actually grounded to the real fundamentals.
• 31:54
  And so this has now become the new hot topic is world models.
• 31:57
  I actually believe Google has already introduced, for public consumption, a version of their world model where you can go into the interface...
• 32:05
  And it's procedurally generating a 3D environment in real time as you move around it.
• 32:11
  And you can choose the POV of like a cat or, you know, a bird flying through the air.
• 32:17
  Yeah.
• 32:17
  And so it's not like a video game where all the levels have been explicitly coded by developers already.
• 32:22
  And then when you get to the edge, you can't walk deeper into the forest because there's an invisible fence in the Pokemon forest.
• 32:28
  Yeah.
• 32:29
  It just continually will procedurally generate, generate on the fly as you move around.
• 32:35
  So as far as I'm aware, that's the first world model kind of thing at scale that's launched.
• 32:42
  But it seems like Jan agrees with this viewpoint and is trying to push forth.
• 32:49
  And with his new company, Advanced Machine Intelligence, they're trying to focus on this idea of building AI world models to actually understand like physical 3D spaces, which becomes much more valuable in applications like AI.
• 33:03
  having any kind of humanoid robotics do anything in a physical environment, whether that's in a manufacturing context, in a residential context.
• 33:15
  Driving's maybe a little bit different because it's such a rules-based system.
• 33:20
  Yeah.
• 33:21
  Whereas having a humanoid robot at home where you want to be able to have it fold your laundry, do the dishes, pick up stuff off the floor.
• 33:28
  It's like unstructured tasks.
• 33:29
  Yeah.
• 33:30
  And I think just in general, too, even for language, right, and like producing text that makes sense or producing recommendations that make sense, having a world waddle is a huge deal, right?
• 33:41
  It'll give you recommendations that actually make sense because it understands how fundamental physics works and so on and so forth, right?
• 33:48
  There's one other sort of...
• 33:50
  interesting palace intrigue corporate drama oh all right aspect to this story which is meta ended up acquiring uh so mark zuckerberg ended up acquiring this sort of ai infrastructure company scale ai which was led by this 20 something you know super genius uh alexander wang not that alexander wang a different one and you know with no e between the d and the r at the
• 34:18
  And they acquired, and Meta acquired their scale AI for $14.3 billion.
• 34:22
  Okay.
• 34:23
  And the idea was it was, they were building like a tagging and training kind of pipeline system that was supposed to accelerate, you know, Olama and all of Meta's AI work.
• 34:33
  Now, he, Alexander Wang was brought in and became Meta's like first chief of AI, right?
• 34:40
  Okay.
• 34:40
  Or head of AI, right?
• 34:41
  And you already have the legend Yann LeCun there.
• 34:44
  Okay.
• 34:44
  Right?
• 34:45
  And so... Yeah, and Jan Lacun is not your head of AI.
• 34:48
  Right.
• 34:49
  And what seems to have been, from the outside looking in, part of the drama here is, you know, Jan is more on the researcher kind of viewpoint.
• 34:59
  So you could probably categorize, like, Sam Altman and Mark Zuckerberg in the scorched earth philosophy.
• 35:04
  Yeah.
• 35:05
  Just spend money, get to the scale fastest, because whoever gets there first wins.
• 35:09
  And then the Jan Lacun and Demis Hesabas are more on the...
• 35:13
  If we have the deepest research bench, we will be the ones who win.
• 35:16
  Yeah.
• 35:17
  And research is the driver of product, not product being the driver of research or product being the driver of implementation.
• 35:24
  And Dario at Anthropic is like kind of somewhere in between those two.
• 35:28
  So it seems to be the case that he was –
• 35:32
  sidelined because he wasn't moving quick enough or Zuckerberg was worried about losing.
• 35:37
  And one of the reasons that Mark Zuckerberg is so motivated to not lose the AI battle against everybody else and why he was one of the first movers to really expect, like do a huge amount of capital expenditure to build out Meta's infrastructure is,
• 35:52
  When we were going through the transition from desktop to mobile in the early 20-teens, Facebook missed the opportunity to build the hardware device for mobile.
• 36:02
  And what ended up happening is they became victim to Apple and Android as it related to getting distribution to their end customers.
• 36:11
  So there was a famous battle between Facebook and Apple where they were like, if you want to keep stealing users' data...
• 36:18
  We're going to kick you out of the app store.
• 36:20
  And Zuckerberg never again, he vowed never again to allow someone to be able to control his distribution in that way.
• 36:28
  That's why he went big on VR and Oculus because he's like, if that's the next medium, no one's going to beat me.
• 36:34
  That's why he went big on AI and all this because if that's the next medium...
• 36:37
  No one is going to beat me.
• 36:39
  However, all Apple has to do to kind of really kill a lot of these, and this is kind of the bare case for a lot of the foundational models.
• 36:45
  If Apple finally decides to do something good and do a local-only, relatively high-quality model on device that no longer needs you to have a ChatGPT subscription or any of these other things...
• 37:02
  and they have billions of devices everywhere and become the first AI of choice, it is going to become very sticky for some of these model providers.
• 37:09
  So we'll see.
• 37:11
  But Jan LeCun is out at Meta.
• 37:14
  He's starting his new thing.
• 37:15
  I think he's going to need more than a billion dollars.
• 37:16
  Yeah, he has a billion dollars already from the likes of Mark Cuban, Eric Schmidt.
• 37:22
  Yep.
• 37:23
  And it values at $3.5 billion.
• 37:25
  Yeah.
• 37:26
  Already.
• 37:27
  Already.
• 37:27
  In like six months of starting.
• 37:29
  Crazy.
• 37:30
  And this is just the valuations of these AI companies.
• 37:33
  It's just outrageous.
• 37:34
  But again, people are viewing it as endgame technology.
• 37:37
  So with that, we are going to wrap up our rundown for this Friday.
• 37:43
  Touched on a nice, fun microbiology story.
• 37:46
  Although it's not that fun if it ends up killing us all.
• 37:49
  Solar storms and earthquakes.
• 37:50
  Maybe this was not so happy and dandy.
• 37:52
  But we had our Are You Smarter Than a Scientist.
• 37:55
  We got 9 out of 10.
• 37:57
  Shared genetics between us and dogs.
• 37:59
  And the battle for AI continues as people throw money everywhere.
• 38:04
  We are going to ask again for feedback.
• 38:07
  This is going to be the third episode that comes out this week.
• 38:10
  We're splitting it out so that they're shorter, more bite-sized, easier to consume and digest on your commute during your workout.
• 38:17
  And it just becomes easier to share with friends if you want them to listen to a specific section.
• 38:22
  So let us know if you do like that structure and that setup.
• 38:25
  We will continue on accordingly.
• 38:27
  I am your host, Lester Nare, joined as always by my co-host and our resident PhD, who may be as smart as a scientist.
• 38:35
  Krishna Chowdhury, we will see you all next week.