StarTalk Radio - Quasar Quirks & Sky Surveys with Matt O’Dowd

Episode Date: June 9, 2026

Did the Milky Way used to be a quasar? On this episode, Neil deGrasse Tyson and comic co-host Chuck Nice explore quasars, the high energy universe, and the movie we’re making of the night’s sky with astrophysicist & host of PBS Space Time, Matt O’Dowd. NOTE: StarTalk+ Patrons can listen to this entire episode commercial-free here:  https://startalkmedia.com/show/quasar-quirks-sky-surveys-with-matt-odowd/ Thanks to our Patrons Alex Nuche, Christian Payne, Gage Ewing, Ryan Whynot, Temirlan, 2 Lives Left, Chad Keeler, Harli Shae Smith, Brad Smith, Norm Bailey, James Peterson, Ryan Coppens, David Whittenberg, Scott Jarboe, Varun Krishnan, Eric Salinas, Mary Seman, Melissa Davis, Stephen Rockwell, Catrina, Max Wilburn, keith Koenigsberg, LEIII, Vincent Loniello, Simon Toth, DoctorWaterGod, Ruthanne Nava, Martineau Alex, Matthew, Phil, Jaden, Arik Drori, Papersneaker, Steven Peeters, Trey Durango, Julianne, Robbie James, Jason Foreman, Liam, Steven Van Vleet, Marilyn, Zakk Why, Ben Wheeldon, Erik Leazure, KONAL SHARMA, Dušan Živanović, Erik Strandberg, berklie novak-stolz, Kazi Mahin Mahfuz, Tim Van Devender, Andrew Martin, Jason F, Charles Joubert, Youcef Kazwiny, Joy Joslyn, Freeman, Jessica, Pat, Phillip Brooks, Michael Hues, Jacqueline Sinclair, Robert Marsh, Botas, Raza Naqvi (Sid), Jake Colón, Christine Bartholomew & Family, Mr Xoot, Dyonté Houston, Daryl, Rob Weiss, Caleb Holmes, Jeffrey Luce, Kellie Owczarczak, and Brandt Reppond for supporting us this week. Subscribe to SiriusXM Podcasts+ to listen to new episodes of StarTalk Radio ad-free and a whole week early.Start a free trial now on Apple Podcasts or by visiting siriusxm.com/podcastsplus. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

Chuck, I love having Mattel Dowd back on. Definitely. Catch us up on quasars, on the Verir Rubin telescope, on big data and AI. Yeah, and I found out that Vera Rubin is actually not a sandwich. Coming up on StarTalk. Welcome to StarTalk. Your place in the universe where science and pop culture collide. StarTalk begins right now.

This is StarTalk. Neil DeGrasse Tyson, your personal astrophysicist. Got with me, Chuckie Nice. Chuckie, baby. What's happening? How you doing? Doing well.

You know what edition of StarTalk this is? What's one would that be? The Matt O'Dowd edition. Oh, that's always good. Matt O'Dowd, welcome back to StarTalk. Such a pleasure to be here. It's my favorite subject. Oh, yes, it would be.

I got you an associate professor up at CUNY, Lehman College. Exactly. Yeah. And that's right. Right across the street from my high school. The Bronx High School of Science, don't I know it? Oh, wow.

It was, yeah. Oh, yeah. Is that at a feeder school to? I think the Bronx High School of Science go to straight to the Ivy League. They go straight to Ivy League. Yeah. Okay. Oh, well.

But some of them do. That's CUNY's loss. Some of them come and hang out. Yeah, it's a good rush. We can hang out there. We've been on to hang out. Okay. And you're a research associate here at the museum.

Exactly. Yes. And you're a host and writer. of one of just the coolest YouTube channels just PBS space time I just so appreciate the work you put into what's on it

how you deliver it and you're just so casually smart casually but I worked very hard I know I just see what I'm saying the effort I see and I feel the effort all behind you just being casually smart in those videos it's like those

$800 rock star haircuts to make it look like you just got out of bed but you just spent 12 hours at the salon exactly that is the perfect metaphor for PBS space time so you have a research specialty in quasars if memory serves

is that correct quasars are a star no pun of my research one or two things involved but yes the quasar is the coolest thing in space I cannot argue with that And also lately you've been into AI and machine learning?

AI and machine learning, exactly. I found that regular eye was insufficient for my needs. Regular eye. Turning to the artificial... I got regular eye. That's where I am. You have sub-regular eyes, sir. At about the Verra-Rubin Telescope, which had first light last year, I think, didn't it?

First light, you know, they kept pushing it. It was actually just a few months ago, the official first light, so it was a little more recent. So it's in the calibration mode at the moment. Okay, so it's an engineering mode. The survey has not yet started. Okay, so the official data-taking survey. Excellent. Okay.

So welcome back. Now, tell me, Quasars, remind us what that is an acronym of? Wow. Okay, so, and there's a whole history, you know. I'm just going to say it's quasi-stellar radio source. Okay. The etymology is disputed,

but quasi-stellar means, so like a star, as in a little pinprick of light on the sky, faint far away, only see it with a telescope. So observationally, it's like a star. Exactly. The stars are not pinpricks of anything.

Indeed. If you travel to one as you have, Neil, then you would know that it's a giant ball of fusion. The quasar is something very different to that, but to us, us mere earthlings looking out, we see these pinpricks of light with telescopes, you can't see them with the naked eye. Some of them also blare radio emissions, so they have these jets, all this cool stuff. And so when these things were first

discovered, there were these pinpricks of light on the sky that were associated with these confusingly loud radio blobs. And so quasi-stellar radio source was a bit of a mouth. Interesting. What's the latest on Quasar? It's because when I was coming up, through graduate school, they were frontier. We were still figuring out. We didn't have the black hole model in place yet. It was contested, not badly contested,

but it was, do we really need a black hole? There's got to be some other way. And they're all far away. How come there's none nearby? So catch us up on Quasars. All right, let me catch you up on Quasars. So, you know, there were the pin bricks, there was the radio, the, like the big, you know,

the watershed moment was realizing that they were far away. when we first took their spectra, we could see that they were moving away from us very quickly, explained by the expanding universe, but they have to be very far away for that. And at those distances, even though they look faint to us, they're insanely bright. So once you calculate how much light there really is,

maybe a thousand times the light of an entire galaxy from a point right in the middle of a galaxy. Wow. And so, you know, people came up with all sorts, swarms of neutron stars, you know, supernova, you know, storms. But the black hole... Any time we're on the frontier, we don't know what's going on.

Right. That opens the floodgates for theorists. Of course. Yeah, yeah. So many papers. But to get a lot of energy out of a very condensed region of space, you know, the black hole is a good way

to do it because there's no way to fit so much energy. Yeah, but if black holes suck, how are you getting energy out of it? That's true. That's true. Stuff gets in, but it doesn't all get in. I mean, so the thing, let me give you a paint a picture of the quasar. So you've got a galaxy with what we call

a super mass of black hole, a million to a billion times the mass of the sun, huge gigantic ball of nothing and gravity. It's a thing of gravity. And so when something happens to drive material too close to it, I mean the Milky Way has one, it's quiet, it's almost invisible. Sometimes they're not invisible. When something happens to drive a bunch of gas, stars, etc., then you end up with this

screaming vortex of material pouring to the ball. You mean a mechanism to move stars? gas and other material from wherever its orbit is down into the center. Yeah, from the galaxy, yeah. It has to drop in somehow. Yeah, like galaxy collisions or like, you know, close, you know, there are ways to do it. Because it wouldn't otherwise have an excuse to be there. You have to find a way.

Something has to move it, right? Something has got to move it. Yeah, just like Earth doesn't fall into the sun, you know, like that, gas doesn't fall into the center of the galaxy like that. It has to be perturbed. Okay, so you have a mechanism. It seems to me that would have been the challenge. How do you get a black hole to release these copious amounts of energy?

Yeah, I mean, it's the energy of falling. And this gas falls a long, long, long way. Can you elaborate on that? The energy of fall. We did a whole thing on this. It's not... You get in an elevator. Right, yeah.

It takes you to a top floor. Right. And then you store up all the energy is potential, and then when you drop, that's energy release. That's the energy that's... That's exactly what he's talking about. Is that what he's talking about? Yes. Oh, so it works.

Well, I think. Hydroelectric dams work by the energy of falling. Okay. The water falls on a turbine. I get it. I get it. Because the black hole has gravity, so there you. I got it now.

Never mind. So the falling, the gas ends up moving at really insane speeds. Forms a whirlpool because things form whirlpools and tries to get in, but the black hole itself is this incredible choke point. It's like trying to cram a galaxy worth of gas into this little point. And so it's screey. into this black hole heats up by friction

at these speeds and that friction liberates so mass is energy etc and mass holds a ton of energy hence nuclear power being so powerful we liberate something like 10% of the rest mass of this in-falling gas is just

pure energy in the form of light photons and so they shine out some of the gas gets in just to just to bring closure to this elevator with on the rooftop, that energy is recovered if you jump, and it becomes kinetic energy. Right.

However, now it's just kinetic energy. How do you turn it into light? Now take me from there. Wow. So you have fast moving gas. Yeah. Something has to now eat that kinetic energy and turn it into light.

So the simplest answer is it's hot. It's searing. It's so it's thermal energy in the end. You've got this whirlpool, these, the gas rubbing against itself and it reaches these insane temperatures. So that right in the middle, your heater is infrared hot. The sun is visible light hot. At the center, this stuff is x-ray hot.

It's just the temperatures are insane. But it's also violent. I mean, it's a vortex of crazy gas boiling into black holes. So you've got these fits and bursts and, you know, energy blasting outwards. And I'm old enough to remember the first X-ray telescope. We were excited because if they found x-rays being emitted from a place where, well, we don't know what else is happening there, it must be a black hole and the gas got so hot

it's now glowing in x-rays. Yeah, and we see those inside our galaxy also on a much smaller scale. The x-ray binaries, which are black holes that are eating their companions start. It sounds very cannibalistic, yeah. It's very cannibalistic, yeah. So we have agreement on this model. Correct?

I mean, the evidence is in, I think. You know, we've now built telescopes that are good enough that we can, you know, for more nearby ones, we can see the gas that in that whirlpool and we can measure its velocities and we can say, well, in order for those velocities, there needs to be this gravitational field and literally nothing but a black hole can produce that gravitation. So is it safe to say that a quasar, I think this is correct, but I've been out of it for so long and I just want to get updated. A quasar is like any other galaxy,

except its black hole in its center is having dinner. Its black hole is in its feeding phase. So the Milky Way black hole has been in that phase before. It's already been down that road. Maybe more than once. So could it be that a galaxy at the edge of our observable universe sees us at the beginning of the universe, because that's the light only just not reaching them,

and they're seeing our supermassive black hole dining upon gases, and could we be a quasar to them? With one exception, I can almost guarantee that. The exception is nomenclature. Quasar is for the brightest ones. The Milky Way would have been a different class of active galactic nucleus, but a quasar-like object, for sure. Okay.

And why would it be, so different because of the size? The size, man, yes. It's the size. The Milky Way's Black Hole, at only 4 million suns in mass, is piddling for a supermassive black hole. But Quasasas have more like a billion.

So they're... Okay, so... They're the big M87, the big elliptical galaxy in the Virgo cluster. That's a hunk of black hole right there in its center. Exactly. And how massive is that?

Is it a billioners around that? Okay, so that's a galaxy that might have been a quasar. And it was. And it still... an active nucleus, but it doesn't have the size of what we call the accretion disk, the whirlpool that it probably once had. But we can still see the jet, so there's like some magnetic fields shooting at the school.

Wait a minute, isn't that the black hole that the telescope imaged? Exactly. What do we call it? The Event Horizon Telescope. That's the black hole. So I'm talking about, oh, we can see the velocities of the gas. No, we have a picture of a black hole now. Yeah.

Nice. Yes. That was banner headlines when it came out. Yeah. tremendous collaboration around the world. Collaboration of people, but also of telescopes. Yes. Because there were radio telescopes literally across the planet

that stitched together their data to be able to get the resolution needed to see that. Very cool. Wild stuff. And this is... Do you know about the data transfer process for the Event Horizon Telescope? No. Plains.

Okay. There was too much data for them to send over the cables. So they had to put them in boxes on planes and send them and stitch it together. Okay, so this is the bandwidth. The internet bandwidth of fax machines were available? Well, exactly. Is that the problem? I mean, we're going to talk about big data.

This is an example of... Yeah, I mean, in my day, I'd be at the telescope, and the worldwide web was in its infancy, and it wasn't even in public yet. I mean, just we had our own channels to get to move data. And there's a point where we had to compare the data rate, transfer from our telescope to our office versus FedEx.

Yeah, no. You load up a tape and FedEx it and reload the tape on the other side. I traveled back from Chile a couple of times with the bag full of tapes of data. You got to think about, oh, make sure they don't go through the x-ray machine and stuff like that. Otherwise, it's gone. Nowadays, it just goes to the archive.

Or you send somebody, you know, the data over the wires, and then it gets there and they're like, it's okay, Timothy showed up. He walked here. And bought us the same information. Bandwidth of Timothy. Yes. Imagine this.

An alien spacecraft lands. An alien comes out, and you are the first human it ever meets. What do you do? What do you say? In my latest book, Take Me to Your Leader, I offer a guide to how to survive that first alien contact, not only scientifically, technologically, culturally, and even socially. not only for yourself, but in that moment,

you are a representative of the entire human species. You want to leave a good impression. So tell me what role gravitational lenses have played in this. And again, I'm old enough to remember the first lenses shown where one had to demonstrate that this curvy little image of a galaxy was the exact spectrum of a curvy image of a galaxy on the other side.

Yeah. That the image had been split. Right. coming around a source of gravity. And we were partying to this. Because just what Einstein had predicted, and so now you guys have taken this to the next level. But I think before you do that,

you might want to say exactly what gravity, even though you just gave, like, if somebody knows what it is, but it's kind of confusing when I first read about gravitational lensing, because I'm thinking of actual lenses and looking through, like, binoculars. This is how you do it. You say, okay, Mr. PBS space time.

Explain that. Okay. You want me to explain stuff? No. All right, so in Einstein's universe, gravity mass bends the paths. It bends the fabric of space, and that changes the path of light. It takes the shortest possible path, which is now curved. And so when a beam of light passes by a strong gravitational field, it arcs around that field.

This is how Einstein's theory was first validated. Arthur Reddington goes to see an eclipse, finds that the positions of the stars behind A solar eclipse. Solar eclipse, exactly. To see the night sky behind the sun and the stars have moved because of the sun's gravitational field. And we see this everywhere now

when we look into the distant universe, we see giant clusters of galaxies and the galaxies behind them have been stretched out like this fun house mirror of gravity bending their light and also multiply image. So you can see the same galaxy

at the bottom of the cluster as at the top of the cluster which you can verify through the spectrum. In the case of quasars, it's particularly cool because the most often scenario is a very different quasar, a more nearby galaxy and the light from that quasar will take two or maybe four different paths around that galaxy.

And so to us, we only know where the light comes from, so it looks like there are two or four images of the same quasar. Was that called the Einstein Cross? The most famous one is called the Einstein Cross. Yeah, because it was a quadrupley-lens object. One of the early ones that was found. And so what you're saying is there's four pictures, and they're the same, because it's the same, the same source is making the four pictures. One key difference is that they're often at different times because the path lengths are different.

So you're seeing the same quasar with offsets of somewhere between hours and weeks. And that's crazy powerful. Wait, wait, hold on. You only know that if something happened on the question. are that you can start the clock. Right. Which, if it's just a static image, you wouldn't know. Exactly.

Exactly. That's fantastic. Isn't it? I mean, that's just freaking fantastic. That means you could predict. If you know, you can say, oh, this burped over here. Right. Watch this right here.

Wait for it. That's amazing. Wait for it. It's almost like you have like a tiny little reset time machine that you're able to observe what's happening, you know? And you say, well, let me see that again. They call gravitational lenses because it's gravity kind of acting as this galaxy-size lens in addition to your telescope, it's a crappy lens because it's made up of stars and, you know, it's not very well ground. And so what you see is a little messy, right? You see

magnification, yes, very helpful, but you also see new fluctuations in the quasar due to the fact that the galaxy itself is made of stars and the stars are moving compared to the galaxy. And so you see all these new effects of a kind of crappy lens. But if you can model all of it, then you can use the gravitational lens to actually map the inner regions of the quasar, which is still very hard because they're still very tiny and very far away.

Wait a minute. You're saying the lens gives you access, deeper access to the quasar than you would otherwise have with just an image of a quasar. Very much. So it's a telescope booster. It's a telescope booster, yeah. Oh, great. Very good, Chuck. Yeah. Cool, man.

It's not... Or enhancer, I would say. Yeah, I mean, you've got to do a little bit of work. You know, it's like when they first put the Hubble mirror up and it was ground wrong and they had to do a lot of work to recover the images. Likewise, these crappy lenses. When you say the word ground, you mean when you take the shape of the mirror, it's... The geometry needs to do that happen? Oh, God.

You don't remember that? They had the wrong shape. Where were you? Where was that? I was like where I was. I was watching Japanese anime. That's where Ible had. Had the wrong shape.

By the way, it had the wrong shape, but it was perfectly ground to the wrong shape. So rather than replace the mirror, we took the other optics and compensated for that other shape and then put it right back in business. Like plastic surgery. Okay. Yeah. Okay.

The reason why that happened, not to get so off topic, is the mirror was tested in situ and it had a perfect shape. And the other lenses were tested in situ, but they were not tested together. Okay, until it was too late. Yeah, still a dumb-ass mistake. Okay.

So, works great. Does that mean now, the farther away the quasar, the more likely you'll have lensing opportunities for things to be in your line of sight? Why, yes, that would make sense. But now it's farther away, but now it's dimmer.

Yeah, I mean, so there's a trade-off. There's a perfect configuration of distance, so the lens, distance to the quasar, and so there's some factors. But it's true that very nearby ones are very unlikely to be lensed. Because there's less stuff, less chance of stuff being between you and it.

And nearby would be, how far away? Uh, is I don't know, downtown New York. No, it's... Nearest quasars, how far out are those? I mean, the nearest ones are, you know, so M51 is one of the newer ones.

Let me take it off of that. When I see maps of the large-scale structure of the universe, the quasars, are the most distant objects in these maps. Is that because they are the most distant, or is it because they're the only things you can see that far away, but other galaxies, ordinary galaxies, are populated among them, and you just can't see them?

Yes. Okay. No, let me, let me, so yes, yes, and one more thing. So first of all, there was a quasar epoch when quasars were, the brightest quasars were the most common, and this is like the middle third of the age of the universe, basically. And we're now post that. There are still some big ones locally. The other thing is that the brightest quasars are quite rare.

And so you just buy statistics, you have to look a long way to see the first one, right? Okay. So if you're in a sparse forest, the nearest tree is likely far away. And lastly, they are the things that we see to the greatest distances. So when you see these surveys and we can't see the galaxies, we see these pinpoints of light.

Right. Oh, there's a galaxy there, but we see it because of the quasar. And you can't see the galaxy because it's not as bright as the quasar. Exactly. Okay. But there are galaxies we wouldn't otherwise seen, were not for the fact that they were far away and lensed. That's also true.

Yeah. Yes. Well, we see galaxies far away now. So you're getting a real assist from the universe itself. Yes. Lensing's very powerful. That's crazy. It's the best.

Predicted by Einstein. Another crumb that just fell off Einstein. He wasn't even thinking about it. He wasn't even thinking about it. He wasn't even thinking about it. He was like, you know what, there might be something called gravitation. Lansing. I don't know that.

That's exactly. F that, who cares? That is exactly how that went down. Wow. Matt, you know that's how that went down. He actually didn't think we would ever observe like gravitational lensing out in space. He thought it would be too weak and too far away.

I think too rare because he was only thinking of it in an exact alignment of two objects. And then you get an Einstein ring. Right. Because if you're slightly off, then it splits the image. It distorts the image. But if it's exactly on it. Then the light is equally likely in any direction coming around it

and you get a ring. I think he figured an exact lineup was rare. But also in 1915 when general relativity was published, we didn't even know that the universe existed outside the Milky Way galaxy. Holy crap. Maybe Einstein did, but he wasn't telling us. Hubble, 15 years later. 29, yeah.

Hubble came along and was just like, yeah, there's a lot more. All right, so thanks for catching me up on quasars. So these sound like complicated problems that need more than just I to solve. Well, I spent my research life staring at these things with my human eyes. But until recently it's been possible. When I started out, we had maybe 100 of these lens quasars. You know, then 300, but we're about to find many thousands. This comes to you from surveys.

It's all going to change with Rubin, which will discover many thousands of these lensed quasars and countless other things and so you know there aren't enough... So is normal eye, normal intelligence not good enough to handle this problem? It's barely good enough to do it even when you have one

because the systems are so complicated. You can't model it easily. You've got these stars in the galaxy moving around. You don't know where the stars are. You have to do that kind of statistically. Within the quasar. Stars moving around within the quasar. Stars moving around

within the lensing galaxy that changes the way the lensing works. Oh, oh. It's very messy. Oh. But it's powerful because those stars, they like sweep across the inner structure of the quasar, like this kind of radar, vum, room, and they can map it. They can map it at the same resolution as the event horizon telescope, but the thousands. Wait, wait, so I, okay, I misunderstood.

You're saying the galaxies that are lensing the quasar, the movement of stars within those galaxies give you varying patterns in the quasar itself. Exactly, yeah. That's what you were saying. That's what I'm saying. Oh my gosh.

Yeah, you have a quasar, you don't really know what's happening in there. Oh, my gosh, that's what you're saying. Right? Right. But I imagine this distant quasar, it's very small, but you've got this lens and you're kind of sweeping

these complicated magnification bands across it. So you see different parts of the quasar change over different times. You can even see when it sweeps across the black hole in principle and see it darken for a little bit. All of this is going to be seen by Rubin. Wow. But.

But. Okay. Like Peeby Herman said, everyone has a big butt. Okay. Okay, go. So our big butt is big data. It's the fact that we need to now model thousands of these things and we could barely do one. And so we are indeed turning to.

AI machine. So not just I, intelligence, right? Artificial intelligence. Artificial, exactly. Which presumes that the artificiality of it is better at it than you are. Yeah, I mean, it's such a catch-all term, and it's not, you know, we're not putting them into chat chagb-teebt where... No, exactly, right. Sophisticated neural networks of different types to do various of these...

I mean, I don't want to sound glib, but it's kind of a matter of just pattern recognition at that point. Right? You're not glib. That's exactly what it is. Okay. These things can be very good at pattern recognition. Better than we are? They can find...

Can they see Jesus in a piece of toast? If you train... But in tortillas, you get them all the time. Oh, if you're just seen a tortilla. Yeah, yeah. The answer is absolutely they're particularly good at that. But, you know, when we come to some data and we look for the patterns, we look for the patterns that we think are going to be there, right?

Or the relationships between the... the parameters that we think are important. With different types of AI model, you can throw in the data, and it will find the patterns, even if they're patterns that weren't expected. But my question, too, is we are excellent pattern-recognizing creatures so good at it that we will see patterns even that aren't there.

Right. So is AI equally susceptible? Is it that good, that it's as bad as we are? Is it so good as bad as we are? That's a sentence. I wanted to stay, but that's it. It's so good, it's bad. So there is that, and there's also the fact that it will find what it's expecting to find,

just like we will. And we see what we expect to see. Do you remember, I mean, it's only a few years ago, but it's now like the ancient era of AI when you train these neural networks to recognize whether something's a cat or a dog. Yes. And you show it a cat, 100% of time it knows it's a cat, a dog, it knows it's a dog. Right. If you show it a chipmunk, it will definitely say it's either a cat or a dog.

Right. Yeah. So it'll see. what, it all depends on how you train it. Just said chipmunk. Okay, we trained it on chipmunks. Okay, we trained it on. So these are some of the challenges.

It's as good as what you put in. Yeah, but if you are training it for what you expect, it's not going to find something that nobody expects. Unless... The serendipity on the frontier of science that we all cherish so highly. But what you could do if I mean, I'm just spitballing here, but you could just allow it to find whatever pattern it wants.

Like, whatever pattern is there. Just find me a pattern. You have to know what pattern means. But it already does. It can extrapolate from the patterns that you already train it on. So then what you can do is... But then it's still extrapolating

from a given pattern. Can you find a pattern for which there is no template? Oh, wow. That's the question. There you know. Okay. So can AI find a pattern on which it has not been trained?

Yes. And that can be an authentic pattern, not something in its imagination. Yeah. Like when we find patterns that aren't there. So there is a power to, you know, in our case it's putting the physics in and training it on what you think the physics is. That's actually quite a powerful approach.

Like this sort of simulation-based. Because it's foundational. Yeah. Yeah. And it tells you what is happening in the context of what you put in. But what if you didn't know what was in there? So there are, so, you know, we talk about supervised learning and unsupervised learning. And there are techniques for unsupervised learning.

supervised learning where you can tell it anything. You just said, what do you see? What do you see? And then it'll tell you something, and then it's your job to interpret what that means and why those patterns emerged. And has the answer ever come back? A bunny rabbit. It's usually chipmunks, actually. Well done, sir. So is that precisely how you're using your AI?

We do a bunch of things. And, you know, Rubin and the other surveys do a bunch of things. In our case, we use variational auto-encoders to take these, in our case, it's the fluctuation over time of these light curves, compact them down into a much more compact space in what we call the latent space, and then use that latent space to try to reproduce the data. In that latent space, we know the patterns are hidden,

and so we have further neural networks to extract things like, what is the mass of the black hole, how fast is it spinning, all of this good stuff, because in principle, the network has learned what the fundamental parameters that went into, generating those fluctuations were. You need to know the physics of everything going on. If you miss some physics, you don't know what you're talking.

Yeah, but you can expand the input physics well beyond what you think is in reality. And we try to do that. We try to say, all right, well, what is the span of all possible physics of these quasars and lenses, et cetera? Let's go much bigger than that to make sure we encompass the true space. And then you can also test how brittle it is, like you can break it and see if it still gives you reasonable answer. So, you know, there's ways, there's ways around the chipmunk problem. You're checking the sensitivity of the system.

Sensitivity and brittleness and see how you... Exactly right, yeah. Okay. That's pretty wild, man. I mean, by the way, I want to say, these are all done by my graduate students in postdocs. Good for you. Are you in Bahamas while they're doing this? Well, Bahamas did come into it, actually. No, let's know the story. I don't need to hear your Bahamas stories. That's fine. The, if anything, I found just the smartest people to do this. And, I mean, these days, I learned programming on Fortran, right?

As did I, I confess. Fortran for me. It's quite difficult for me to really grok the ins and outs of all of this. Right. So I just try to. You an old fart. It doesn't make a difference. You already did your time.

I did my time. Yeah, you did your time. You did your time. I get to nod, wisely, and not breathes, right, and take credit for their work. That's how it goes. And one day they'll get to do that to the next generation. That's how it goes. You know what I mean?

It's beautiful, really. So we've seen photos of the Veraruban telescope in the Andes Mountains of Chile, where we have a lot of telescopes. And there's the dome, and then there's this whole other section sticking out the side. The shoe, yeah. Oh, I thought about it that. To me, it always looked like a shoe. It looks like a shoe sticking out because it has a rounded front. I presume that's where all the data, the big data is happening.

Yeah, I think they stack the data in there. I think there's a lot of... And then they fly it up to whatever. There's a lot of on-site processing because you can't just ship it all. So is it because this telescope is uniquely in need of data processing support that it was conceived and designed this way? Well, there's also, you know, there are people there doing... important engineering things.

They haven't been replaced by AI like we have. So engineering support, but also the computing facilities that are there. And yeah, I've never been. I want to go and take a tour. Of the shoe. Of the shoe. Yeah, yeah. So what it seems to me, I don't want to speak for you, but tell me if I'm correct,

that most people's fear of AI is that it'll take their job. Whereas when you're a scientist on the... the frontier, such as yourself, the AI allows you to step where you could not have stepped at all. So it's not the same thing as replacing a job. It is empowering you to think more creatively about your thoughts on the scientific frontier. Is that a fair characterization? It's such early days in this revolution that it's hard to say where it's going to land.

Right now, it's insanely powerful. in many, many respects. It takes away work that, you know, we didn't want to be doing anyway. It's a lot of great work. It's incredibly powerful. But it's also, you know, the new reasoning models are able to do things that previously graduate students were doing. And the hope, the hope is, oh, well, now graduate students can be freed up to do better things.

Yeah. More creative things. That's the hope. The reality might be that the AI is like, look at you, dumb ass. That's what you're just tough. I can't believe you thought that this was something viable. God, who hired this dude?

That's a little scary. I mean, on the other hand, the stuff that graduate students used to have to do, which is stare at this boring data forever. Right. You know, an AGI, artificial graduate student intelligence. Very good. Can now. That was nice.

do that. They can be occupied. Essentially. And so the hope is that the professors won't say, oh, I don't need graduate students anymore. They'll say, oh, graduate students. Now do this. I don't get this stuff.

Please do this. Now do this. Now let me ask you this. Is there any benefit to the graduate student doing the grunt work? Does that, there's something that can come out of that for our brains? I'm going to say no. Here's why.

Really? In my day. And your day. Okay. Pre-AI. Okay. But computer power was growing exponentially. Mm-hmm.

There used to be a course in graduate school on spherical trigonometry. Which nobody needs now. Because the computer does it. Exactly. Okay. Spherical trigonometry, you know, trigonometry normally on a flat piece of paper. But on the dome, you have angles between stars and moving the telescopes and what's the shortest slew path between two. That was, that's all spherical trigonometry.

Gone. We just push a button. And it's done. Telescope calculates it. Now let me ask you this. You took spherical. But you took it, right? No, no, no.

It was like two years before I got there. We stopped teaching. And do you understand spherical trigonomy? No. That's the real question. I mean, the question is, are there intuition that? No, I can. We miss now.

I think there could be intuition. It's an intuition thing. Yeah. That's really what I'm talking about. Going through the wax on wax off of graduate school. Right. Oh, what were you learning? I don't know, but now I know Kung Fu.

We don't know what we lose, I guess. Interesting. So I, okay, let me give a counter story to that. All right. When I was in graduate school, a member of our department, a faculty member, was world's expert on galaxy classification. Okay. World's expert. Okay.

There's never been an expert such as him either before or since. Okay. Okay. And so I'm in one of his classes. We are classifying galaxies. Okay? It's like, this is stupid. Why am I doing this?

All right. Okay. And nowadays, computers classify galaxies. You don't need to do this. But when I look at a galaxy that you just took a picture of, I have a whole other relationship with it that you don't. I'm feeling it. Because I'd look at hundreds and hundreds of these, maybe thousands of these.

And so it's in me in a different way. It's almost a muscle memory of... what it is and why it looks that way and what I could. And to what I was saying earlier, who knows what inspiration that is inspired by just that reservoir of seemingly useless knowledge that you were holding. Like he said, the wax on, wax off. Yeah, I mean, precisely the... And, you know, knowing what's under the hood, knowing how the sausage is made, you know,

these days, future generations of graduate students won't, know how to code because they talk to the computer, they vibe code. A but will code. And they never make an if statement. And what do we lose? Maybe nothing. But I feel like there's something about knowing what's under the hood. That kind of helps.

Well, it helps you know what the true capabilities are. Do you know how you're smart phone works? No, you don't. Yeah. Like you don't know what the vulnerabilities are if it's just a black box, I guess. But if it's a perfect black box, there are no vulnerability. Okay. So we just need to build.

bet. Hey, this is Kevin the Somelier, and I support StarTalk on Patreon. You're listening to StarTalk with Neil deGrasse Tyson. Let me just bring closure to this data challenge that we now have. We are awash in data. The data rate for the Rubin telescope will exceed that of any previous telescope in our portfolio. Orders of magnitude.

By orders of magnitude. Wow. I'm reminded of what was the first time they turned on the telescope and they discovered like a thousand new asteroids crossing the sky. Right, yeah. And you could do that because it's taking its repeated imagery, which is the only way you'll know if something moves.

Because otherwise it's a still frame. It's a static shot. And is that a star or is it something that goes bump in the night? So that forced us our field, our community to innovate in ways we didn't have to before, right? So it was all good. It headed in the right direction there. Yeah.

Do you mean as astronomers or as a conversation? All of the above. Oh, that's great. You were so incredulous. You're like, yeah. You offended. No, Neo. I'm wasting my life.

It's funny because it's true. I only say that because we had needs for imaging the universe that exceeded what Kodak, now many decades ago, routinely produced. And so they put a special team on our needs and created special emulsions that were more sensitive, that were larger, and so when CCDs came out, we were the first to fully exploit them.

And before anyone even knew what they were. And poor Kodak did not take advantage of that. Kodak looks at it. It's just a flash in the pan. They're like, we're going to do this for these nerds. People love film. Yeah, these eggheads need these great feats. They'll love film all the time.

So only when it became a commodity did, were we no longer the leading edge of that. But it seems now, with this level of data, is this more data than anyone has had to think about before? Or are advertisers mining data off of social media accounts? And is that a greater, a greater repository of data? for them to sell this product. I mean, it's greater for them. In terms of bits, I don't know if it's greater. So it's enormous.

It's, you know, for the first time taking a full image of the entire Southern Sky every three nights for 10 years. Wow. I can give you some numbers about how much data that is. Like the CCD is enormous. It takes like 400 HD TVs to show one image. 400. 400.

And that's one image, and the southern sky is like 3,000, 4,000 of those images. Just to be clear, the Hubble telescope field of view is a fraction the size of the full moon. Wow. And this is like 40 times the size for one of those images. Of the full moon. Yes. That's amazing. So that's how you can.

So if you said Hubble, give me an image of the whole sky. Okay. Call me in 30 years when I finally. That's right. Because I'm going to have to stitch this all together. They're going to mosaic this. Right. And so.

Call me in three days. It's pretty fast for the whole sky. Wow, that's the very rule of it. Okay. But like you said, it's a movie of the sky. So we see things changing, things going bump in the night. We see the quasars flickering at the edge of the universe. All of it.

And it's all just going to be mainlined for 10 years. And what do we do with it? Well, we work very hard. You guys are making like a flip book of the universe. It's a flip book of the universe. It's a flip book of the universe. It really is. Holy crap.

big flipbook. It's heart of the pages. It's like several football fields. You put the thumb across the page edges. Yeah, that's funny. I got a paper cut. Yeah, I lost my hand. So just to anchor this in proper context at the risk of repeating myself, in my day, you go to the telescope, you take a picture and you take the picture home and analyze it.

If something moved, you have no idea. You have no idea. And it was not that big a problem because, you know, Stars live 10 billion years, five billion years, and you're there getting a 30-minute exposure of it. You're not expecting fireworks. Right. In that moment you're looking at it. But maybe there are.

Somewhere in the universe. Well, there are. We know there are. We know a lot of what the fireworks are, but there's a lot that we don't know also. The universe is pretty violent. It's quite dynamic. Yeah. It's tough.

Yeah. So it'd be a shame for us to sit this close to each other. and not compare notes. So you've got a YouTube channel. You bring delivered content that's fun and interesting and exciting to hear. And it's a mix of not only what is frontier science, but also it's fun science.

So it's clear that you're doing some cherry-picking of what you could be talking about. Plus, you look like the sexy professor that one might daydream about. Well, thank you. The other thing that we go. So share with me some of your tools and tactics that you've found most potent in your efforts to bring the universe down to Earth. You know, so PBS Spacetime feels a particular niche, I think, which is that we do go hard.

Like we've covered the holographic universe and quantum mechanics and all the way to the edge of the holographic universe as, you know, as well as, you know, more traditional space stuff. So we found very early that there's a huge appetite for kind of seeing under the hood, like how science happened, how scientists actually talk. And I think for the longest time, people have felt a little bit babied by a lot of popular science media. And they know when they're being talked down to. They know when they're being talked down to. And so I think one thing that I do well is I have a good jargon detector. I know when something's jargon.

And, you know, jargon doesn't have to be like a specialized word. It can even be a specialized use of a word. And so the point is that so much of science, even the stuff that's hard, is accessible to human language. And it can be talked about in human language. And I think that scientists are not the best people

are doing that because they talk professionally. They talk in professional. You're not trained for that. It's all shop talk for scientists. It's all shop talk. Yeah, but all professions have shop talk. Exactly. Right. So what's different about science?

Science is hyper-specialized just because it's old. We know so much about the world that to make any progress, we have to dig deep and narrow to make any progress. And so the language around each subfield tends to be very specialized. And every time you get a subfield over subfield over subfield. It's excluding of others because of the language and because it's, you know, it attracts a certain type of people, nerds. You know, it's true.

And they enjoy going granular on information. Yeah. So I think over time, science has become, I kind of think of it as genre-fied. So, you know, something is sciencey if it's hyper-focused and detail-oriented when really it's about curiosity about the world. Interesting.

So I think the beginning of science and what it really is is just a, it's curiosity with, you know, organization. And this is where we are now after centuries and millennia of doing this. We know a lot, right? But a side effect of that is the siloing of fields between each other and between scientists and non-scientists.

And just the fact that the scientist and science feels like an other type of thing. It's a certain type of person who becomes a scientist and is into science. Whereas, you know, one thing we try to do is to make things sound a little more collective. You know, I rarely say... Inclusive. Inclusive and collective. in the sense that, you know, yes, we should be very proud of Albert Einstein for coming up with general relativity, but we should be proud of humanity also for coming up with, you know, for one of us for figuring this stuff out.

You know, we figure this out. I think about that all the time. That's why I have an active disinterest in my genealogy. Because I want to be what I want to be based on what I know humans are capable of. Right. Not based on Uncle Fred. somebody before you. Right. Right.

But also it's lazy when people do that. It's completely lazy. Very lazy because what you're saying, it's like when people say, we did it. We won the world theory. Who's we? And I'm like, really? What now?

When's your contract? Like, shut the hell up. No, no, but you're allowed to participate as a species in the achievements of your species. Yes. I think you're allowed to do that. Without a doubt, but the fact is that... I'm going to say we figured out how to build a system. suspension bridge and figure out to go to the moon.

Yes. And I don't want you coming behind me say, well, what part of the project did you work on? Well, no, there's a different. That's different, though. And here's why. Because this is my tax money that went in? Not even that. You did.

Your tax money did help build that. But the thing is that science is not so specialized like a sport or something else like that, that you can't do it. You can actually do it. If you want, you can understand, and believe me, I'm speaking from experience. You can understand this stuff. It's a little difficult. It takes a little bit of work. But once you do that, it's like, oh my God.

But I think the public sense is the opposite. I think the public sense is that sport is more accessible than science. And that's my point. But my point is this. Sport is not more accessible. Of course. You can never hit a home run in a major league park. I'll give a damn what you think.

you can do, it's never going to happen. So it's like the person graduating high school saying, or college saying, I want to be a professional basketball player, let's say. And they're way more neurosurgeons or aerospace engineers. Exactly. So that's too hard. I want to do this easy route. And it's the hardest thing in the world.

You're never going to do it. But the thing is kids play basketball and it's not, you're not really into, so you're not genre-fied for liking basketball, you're a kid. Right. And you play it, not because you might one day make the NBA, but because it's fun. But science doesn't have that nearly as much. Right. The kids who do science, they're the nerds.

But you guys are trying to do science. I think you're not giving yourselves enough credit here, okay? What you do is make science like basketball. And quite frankly, there's an audience out there that really feels that. Yes. And they appreciated greatly. Yeah. Well, thank you, Chuck.

I feel like that is a big part of the goal. I think we have a lot of work to do. Can I tell you my transition? Thank you for saying that. That was insightful. Oh, cool. Listen, even a broken clock is right twice. Okay, broken clock.

Okay. Okay, so in my sort of rise in visibility, in the early days, people would come to say, are you, Neil Tyson? I said, yes, I am. Okay, tell me about black holes and what you said the other day. So all it was was a food for them. The fame was the fact that I had excited their curiosity.

They didn't care what my favorite color was. As my visibly got higher and higher, more and more people would just say, oh, can I get your autograph at pre-selfsearch? Can I get your autograph? And wouldn't it? Don't you want to ask me about the universe? No, I just wanted your autograph. So I felt cheap.

When was the last time you told someone about black holes? I mean, everyone you tell about black holes. So this continues, and more and more people, now they want to take selfies, that's fine. And I oblige. But I felt incomplete by this until someone told me. And I said, Neil, do you realize you're a scientist and people want your autograph?

Right. You're a scientist, and people want your autograph? to take a selfie with you, like a rock star or athlete? And so I had to, real, because they're not going to go to the athlete and say, please explain your, they just want to, they want to be, they want to connect with the athlete. They just want to connect in that one way. And so I realized, oh my gosh, the science is sharing the space that it only previously been occupied by- And it's happening more and more. By the way, I got to,

Alex, our producer, you were not here. I've never said this to you, but I'll say it now because he was here. And we were talking about you. And I said, we were talking about you. And I said to Alex, do you know how hard it is to be world famous? Science. What? Am I lying at, Alex?

Yeah. Yeah, there might be just a few dozen scientists who became... It's what I'm saying. But what's great is I love what you said about the accessibility and not talking down and allowing people to find the wonder in science. Because I think that not only is it necessary for science to continue to thrive, socially. But I think it's imperative

for democracy, for our species, for the future of not just the country, but the world with respect to the things that information that we know that will save us as a species, like

climate, you know, and all these things really are very deeply connected to science and scientific literacy. And that's the truly important thing that you guys are doing. He's running for office soon. No, no, I'm sorry. You got my vote.

But it's not just what science can do for us. It's this suite of modes of thinking and tools of thought and just ways to apply curiosity and rationality. Absolutely. That are absolutely not restricted to what we now think of a science. It's just thinking. It's just thinking carefully. Absolutely. And it's so important in all works.

A state of mind, a state of thought, a state of curiosity. I love it. Yeah. The last thing I'll say here is I had to. learn this after I stopped teaching classrooms and started writing books and doing podcast. There's no obligation in a person's first encounter with your expertise for them to learn everything on your syllabus. Throw out the stuff that's boring or uninteresting.

Yeah, it's got to be there in a classroom because you need the sequencing. But pick the stuff that's really cool. Teach them that. Then they say, this is cool. I want to learn more. And then they're at a level where they can get the nuances that. you left out in the first place. Yeah, yeah.

But also, like, throwing in a few, like, little deep views is nice also. Yeah. Oh, these are, no. Teasers for the future. Oh, yeah, yeah, yeah, yeah. But I'm just saying often there's detail that's just simply unnecessary. Sure. Yeah.

At the first pass. And it's hard to, it's hard to know how to spot that and excise it. We're very precious about our details, us nerds. Yeah, it is. It is. Dude, we got to end it there. Yes. Oh, yeah.

Matt. This was fun. Thanks, Matt. Always good, having Matt on the show. It's nice to come down to the fifth floor and chat about quazos. All right. Chuck, it's good to have you, man. Always a pleasure.

And your special is still, your comedy special, still airing on our YouTube channel. And it's called, what? Chuck Nice. Just Smart enough. Because I'm sitting between these two. So that's about as good as it's going to get for me. I love that. I think you're smarter than Just Smart Enough.

But, all right, this has been another installment of StarTalk. The Matt O'Dowd edition. Yes. Until next time, keep looking up.