Into the Impossible With Brian Keating - Do We Have Free Will and Will AI Have It Too? (ft. Sabine Hossenfelder)

Episode Date: August 14, 2025

Please join my mailing list here 👉 https://briankeating.com/yt to win a meteorite 💥 Does free will exist? It's a question that's haunted philosophers for centuries. But physicist Sabine Hossenfelder has a provocative answer that might just disturb you. She says free will doesn't exist. Everything is determined by the laws of physics. But here's the paradox that's fascinated me. I've talked to Sam Harris, Dan Dennett, Robert Sapolsky, all brilliant minds who agree that free will is an illusion. Yet, when I asked them if they've ever encountered someone who acts like they don't have free will, they all said no. So how can this be? How can we all be determined by physics, yet live as if we're making genuine choices? This isn't some abstract philosophical exercise, no. Sabine's insight has real implications for artificial intelligence, quantum computing, and the future of consciousness itself. She explains why AI may already have the same kind of agency that we do, and why quantum computers could reveal that we fundamentally misunderstand the nature of reality. Key Takeaways: 00:26–04:39 – Determinism over “free will,” reframes self as information processor. 07:21–09:19 – Defines agency as internal deliberation vs. external input; AI has low agency. 10:48–14:50 – AI can learn without bodies; emotions could emerge via reward functions. 15:18–17:16 – UFO claims are mostly flaky but worth some serious consideration. 18:16–22:52 – “Quiz with It” uses AI for quizzes; AI may replace poor lecturers. 24:37–30:37 – Warns of LLM “lock-in”; expects new AI paradigms. 31:15–32:42 – Quantum computing promising for finance optimization; uncertain impact. 34:15–36:37 – Likes debates; thinks “theories of everything” overemphasized. 39:09–42:14 – Would fund quantum gravity & measurement problem experiments. 43:21–50:31 – Supports fuzzy dark matter search; inflation underdetermined; MOND captures some truth. 51:16–57:26 – Skeptical of DESI’s dark energy claim; cosmology data messy, model-dependent. 58:40–01:00:14 – Academic decline due to systemic incentives. 01:00:50–01:02:36 – Quantum tech advances reviving foundational physics interest. 01:09:30–01:12:23 – AI-generated junk (“AI slop”) threatens science publishing. ----------------------------------------------------------------------------------------- Additional resources: Check out Sabine’s channel: https://www.youtube.com/@SabineHossenfelder ----------------------------------------------------------------------------------------- Join this channel to get access to perks like monthly Office Hours: https://www.youtube.com/channel/UCmXH_moPhfkqCk6S3b9RWuw/join 📚 Get a copy of my books: Think Like a Nobel Prize Winner, with life changing interviews with 9 Nobel Prizewinners: https://a.co/d/03ezQFu My tell-all cosmic memoir Losing the Nobel Prize: http://amzn.to/2sa5UpA The first-ever audiobook from Galileo: Dialogue Concerning the Two Chief World Systems: Ptolemaic and Copernican https://a.co/d/iZPi9Un 📺 Watch my most popular videos:📺 Neil Turok https://www.youtube.com/watch?v=Dt5cFLN65fI Frank Wilczek https://youtu.be/3z8RqKMQHe0?sub_confirmation=1 Eric Weinstein vs. Stephen Wolfram https://www.youtube.com/watch?v=OI0AZ4Y4Ip4?sub_confirmation=1 Sir Roger Penrose: https://youtu.be/AMuqyAvX7Wo Sabine Hossenfelder: https://youtu.be/g00ilS6tBvs Avi Loeb: https://youtu.be/N9lUceHsLRw Follow me to ask questions of my guests: 🏄‍♂️ Twitter: https://twitter.com/DrBrianKeating 🔔 Subscribe https://www.youtube.com/DrBrianKeating?sub_confirmation=1 📝 Join my mailing list; just click here http://briankeating.com/list ✍️ Detailed Blog posts here: https://briankeating.com/blog 🎙️ Listen on audio-only platforms: https://briankeating.com/podcast #universe #podcast #briankeating #intotheimpossible #science #astronomy #cosmology #cosmicmicrowavebackground #intotheimpossible #briankeating #sabinehossenfelder Learn more about your ad choices. Visit megaphone.fm/adchoices

You said this place was steps from the water. We just haven't found the steps yet. How much did we save? Enough. Enough to get lost. Or you could book a stay with Hilton. Welcome to your ocean front room. Just steps from the water.

The Hilton sale is on now. Book on Hilton.com or the Hilton app and save up to 20% to get the stay you expected. When you want savings, not surprises. It matters where you stay. Hilton for the stay. Study And play.

Come together on a Windows 11 PC. And for a limited time, college students get the best of both worlds. Get the Unreal College deal, everything you need to study and play with select Windows 11 PCs. Eligible students get a year of Microsoft 365 premium

and a year of Xbox GamePass Ultimate with a custom color Xbox wireless controller. Learn more at Windows.com slash student offer. While supplies last, ends June 30th, terms at AKA.m.S. slash college PC. Does free will exist? It's a question that's haunted philosophers for centuries.

But physicist Sabina Hansenfelder has a provocative answer that might just disturb you. She says free will doesn't exist. People just go, what do you mean by free will? And I was trying to say, what doesn't really matter? Because what I'm talking about is like, what do we know about the laws of physics? And why is it so hard to make sense of this, like on a personal level? If you know that your brain is made of particles and we know, the laws that those particles are ruled by.

Everything is determined by the laws of physics, but here's the paradox that's fascinating me. I've talked to Sam Harris, Dan Dennett, Robert Zupolski, all brilliant minds who agree that free will is an illusion. Yet, when I ask them, if they've ever encountered someone who acts like they don't have free will, they all said no. How can we all be determined by physics yet live as if we're making genuine choices? This isn't some abstract philosophical exercise, no.

Sabina's Insight has real implications for us. artificial intelligence, quantum computing, in the future of consciousness. She explains why AI may already have the same kind of agency that we do, and why quantum computers could reveal that we fundamentally misunderstand the nature of reality. Savina, it's so great to see you again after more than a year. I don't think we talked very much in 2024, so it's great to see you. Good to see you, Brian. You've had a wonderful video yesterday day before about this thought that changed your life.

And to me, I was a little depressed because it seemed to say that, you know, everything is physics and everything is determined. And therefore, you know, you should maybe naively not be happy, but you turned out to be very happy. So can you explain what determinism is? Because Sabina, I've talked to everybody. I've talked to Sam Harris. I've talked to you. I've talked to Dan Dennett, the late Dan Dennett, his final podcast. I've talked to Robert Sapolsky.

They all say free will doesn't exist. And yet, I say, have you ever met somebody who acts like they believe that they don't have free will? And they all say no. So how can it be that we don't have free will, but everybody acts like we have free will? Well, it all depends on what you mean by free will. So if you watch my video, you'll have noticed that I avoided using the word. This is why I was talking about determinism, because otherwise, people just go, what do you mean by free will? And I was trying to say, well, it doesn't really matter. Because what I'm talking about is like, what do we know about the laws of physics? And why is it so hard to make sense of this, like on a personal level, if you know that your brain is made of

particles and we know the laws that those particles are ruled by, right? So where's my place on this? Now, you said, I seem to be happy about it, but really it was more like, I was unhappy before. So I was trying to solve a problem and I finally succeeded. So what I talked about in the video was that I finally eventually managed to shift my point of view on myself, basically, and made peace with knowing that everything is determined up to this random element. And I read Sam Harris' book, which, you know, it's only like 60 pages or something. And I agreed with everything except for the one paragraph you wrote about physics. But, you know, but, you Yeah, otherwise, you know, I think he's right.

I think what the disagreement mostly is about is what people even mean by free will, which is exactly why I don't like to talk about it. I just think it's a nonsense term. Like even the combination of free and will doesn't really make sense. So if you think your will determines your decisions, then it's not free. Like that's the entire point of having a will that you make the decision, right? So what I was trying to get at in the video is that I began to think of myself as an information processing thing, basically, that goes through the world. And I think it's made me more aware the information that I consume and how much it affects me.

So that was my point. And so, of course, this entire question of free will is becoming more and more relevant with artificial. intelligence that is rapidly coming into the world more and more powerful. And people starting to ask, like, does it have free will? At which point do we agree that it has free will and is worth protecting on some level? Sam Harris says it will have free will or already does. So how do you react to that? I mean, how can you say that, you know, free will as a concept doesn't exist when we think of AI at least as, you know, for now, as, you know, as running on G. DPUs and with large language models with billions of tokens.

How can it be possible that it could be even said to have that? If we're determined, then certainly the determiner of the programmer of the AI must be in some super, you know, superposition, you know, super above the AI level. Do you agree with him, first of all, that AI can be said to have free will? Modula the fact that we can't define it. Yeah, well, I'd say AI doesn't have free will exactly the same way that we don't have. have free will because in both cases it's determined. But I think there is still at this point a measurable degree of difference. So I think the most compatible definition of free will that I know is the idea that it's got something to do with agency. You have a system, how much of the

behavior of this system is determined by the external inputs and how much of it comes from some internal, complicated deliberation about what's going on. And I think it's, you know, even if you can't, you know, rigorously quantify it, like, I mean, you know, philosophers, right? It's not like they write down an equation and then say, you can calculate your amount of free will using this formula. I mean, that would be nice, but we're not quite there yet. But so even on this vague level, you can say, yeah, I think humans have more free will

than a phone. And AI at the moment is somewhere in between, right? It's clearly doing more than a phone where you push a button and it pops up a window or something. It's not that simple. But it's also much more constrained than humans because it's still very apparent what kind of prompt they're working off. They're not really doing their own thing. So they're doing a few of those things, which is why philosophers, I think, are arguing over it. Like with AI agents coming online, they're setting their own goals and they're, you know, producing a plan. And then they work off this plan and they think about the plan, you know, then they revise it. And it's going to get more complicated along those lines.

And so I think it's a valid question to ask, but personally, I'd say like it's a very, very low level of free will. I wonder if you could say in German for me what Einstein said in 1907. He said the following was his happiest thought that here he is on the finger puppet level. He said if he was in free fall that he would experience no gravitational force. Can you say it was the happiest thought of my life? Because that's what he said. That's what he called that realization, the Einstein equivalence. Can you say it was the happiest thought of my life in German?

This was a glucklisting a goodank of mine as liams. So now it might be that I mistranslated back something that was mistranslated into English in the first place. So I'm not exactly sure that this is the right quote, but he said something along those lines for sure. Yeah. And so you want me to explain the elevator experiment or thought experiment? Not really. I want to pivot to what would it mean for a computer or an AI to have a happy thought? thought. And furthermore, could an AI have such a thought because it doesn't have a body?

And so we wouldn't know what that sensation of freefall that Einstein visualized and felt viscerally. How can a computer have a happy thought and how could it visualize what freefall feels like in order to replicate and be an AIAE, you know, an artificial intelligent Albert Einstein? Yeah. Well, that's a very good point. So like, to ask this question, will AI ever become truly intelligent without actually having a body? Like, do you actually need to be? put it into a robot. There are two answers that I have to this. One thing is you can simulate a robot walking around in a virtual reality. And so it doesn't actually need a physical robot just like a simulation of it, if you see what I mean. And people are already doing this.

Like this is the thing. Like if you want to, if you want to train your robot, you don't let it walk stairs up and down for a billion years. you squeeze up billion years into a few days on a supercomputer or something and I let it learn to walk stairs in a virtual reality. So that's already a thing. The other thing is that I generally, I believe it to be possible. Like you could reach human level intelligence with the lucky thoughts and, you know, genuine insights about reality without the direct.

sensory input from physical reality, but I believe it would take a very long time. Because if you think about it, like it also took humans a very long time to become as intelligence as we are now. Right. So this is what you want to avoid. You want to get that fast, right? And so this is why I believe that the most likely way we will reach human level intelligence is by some sort of access to the real world so that they can learn what it actually means

to interact with something. Or if you want to use the causal model approach, you need to be able to interfere with certain causal relations to figure out how they work. And we basically do this from the day we're born or even before that. Who knows? And so I think this is going to be a real game change. You know, once we have robots walking around, they're going to collect their own data. They're going to create their own models of the world, but much more efficiently than we could have done.

You also asked about happiness, like can I always have an emotion? Well, again, at the moment, I think the answer is no. not in any reasonable meaning of the word. But you always have to ask like, what do we actually mean by this? So for us, happiness evolved as a reward function, right? It's the response to doing something right. So you got a good meal. That's good.

It's going to help you survive. Dopamine. Yeah, the dopamine hit. We have to have robotomene or something for the robots to have. Right. And of course, AI already has a reward function, right? So it has the root of it because it needs that to learn. And the newer models are also learning to improve themselves by way of a proper reward function. And I think at some point, that would be very similar to a particular emotion, in this case, happiness. And then you can ask, well, what about all the other emotions? And then, you can ask, well, what about all the other emotions? And then, You know, we have to talk about like, will they actually have different sort of? Yeah, pain. Right.

And will they have different emotions? Like, will AI have emotions that we don't have that we can't even relate to? I don't know. I have no idea. Like, it's also, you could also ask this question about aliens, right? So if aliens have emotions, that they have the same emotions, that they have different emotions that we won't be able to relate to? Well, who knows?

I've no idea. What do you make of all the alien, you know, kind of sightings that are in the, the zeitgeist, as you might say. What do you make of these claims by respected, you know, military personnel, mostly in America, by the way, but what do you make of it as a physicist? I mean, wouldn't you be happy if this were true and you could learn the laws of the 23rd century today? Totally. Like, I would love to see intelligent life from other planets within my lifetime because they've got to have, like, if they manage to travel here, they must have more advanced technology than we have.

They could probably help us solve many problems, answer many questions, and so on and so forth. I was just chatting about this with Kurt the other day. It's a difficult topic because I think that a lot of what people talk about is incredibly flaky and its nonsense. And on the other hand, I think we shouldn't just dismiss everything. because if we make a mistake with that, it's going to be a huge mistake. And I think it's, you know, I think it's actually possible that there's intelligent life out there and it sends some probes into the galaxy like we are planning to do.

And maybe one of those arrived here and it fell into the ocean. You know, that seems totally possible to me. That doesn't mean that I actually believe we found one. So I'm a little bit unhappy with the situation that we're in in which everyone who talks about UFOs or UAPs or whatever you want to call it now, you know, it kind of moves themselves into that crackpot corner because I think we need to take this more seriously. Also, I have to say, you know, if you listen to any of these Congress or Senate debates that they did, there's this one guy from NASA. And the guy from NASA, like every time he basically, he gets to say something, he basically says,

oh, and by the way, we need more funding. That's right. He knows the, he got the assignment, right? And actually, speaking of assignments, you have this wonderful product or, you know, service. Actually, it's free called Quiz With It, right? I want to ask you about AI education. First of all, do you use AI in Quiz With It? Is there going to be a job for people like me in 40 years? professor, you know, standing up in front of a bunch of students with a piece of rock and scraping

on another piece of rock and then they're listening and or is this just going to be my avatar talking to their avatar or better yet, you're a better explainer than I am. Why can't they take quantum mechanics with you instead of with me? Or how about they take it with Borr, Schrodinger, you know, Heisenberg, et cetera, or learn GR from, you know, good old Einstein himself. So what's the future of education in the age of age of age? It's peak power. It's peak. pollination season, and my business is scaling fast. To keep the nectar flowing, I need a phone plan with top priority data speed. That's why I chose GoogleFi Wireless.

My connections stay strong even when the hive is buzzing. Plus, unlimited plans started $35 a month. Now, that's a deal that doesn't stay. Explore Google Fi Wireless plans today. Plus taxes and government fees. Google Fi Wireless is not subject to data traffic deprioritization during times of high network usage. AI. Will professors like me have a job?

Please say yes. And how do you use Quiz With It? How does Quiz With It use AI? And give a little plug or talk about a quiz with it, please, in case my audience. Okay. Yeah. So Quiz With Aid is a platform that I developed together with a team of software engineers. And yes, we do use AI.

We have a GPT link. And so the way that it works is that it'll automatically create questions with answers on any topic that you enter. So you can just enter the topic and then you'll get questions with answers. Or you can be more specific about it and you can actually enter some text of some, I don't know, blog post or something. You can also just enter the link and we'll pull down the text automatically. Or you can, and this is what I do, you can use YouTube video and we'll pull down the transcript

and then we'll generate questions from this and you can quiz yourself on it whether you actually understood it, whether you remember it correctly. And so the reason I did this was that, I mean, there are lots of quiz ups out there, but none of them actually did what I wanted to do. I wanted users to be able to collect points from taking multiple of my quizzes and also for other creators to be able to do the same. So that was basically the idea. Now, there's this broader question of what's AI going to do to education. And I think that's a really big topic at the moment because that's really a sector where AI can have a huge impact

and it's already having an impact. And the question is like, where will it go? I mean, at the moment, it's very hard to tell. I think in the end, it'll come down to how much do we value human interaction, right? So, like, I mean, so this example you had like, can students learn physics from Einstein or Bohr? I think the answer is probably no, because we don't have enough material to train the AIs on. Like, there are only so many recordings from a century ago. Oh, but we have, I mean, Einstein wrote a million, two million words. I mean,

I have Galileo's, all of his writings on the dialogue, which is about 800,000 words. We recorded an audiobook with Carlo Revelli and Frank Wilcheck and others. There's plenty of written material, right? We'll never get Galileo's voice, but I don't think we need it. Yeah, but when these people write their academic papers, that's not like 100% who they really were. So like if you write a paper and then you were to talk about it, it's not exactly the same. This is what I mean. So we can try to reconstruct an approximation to who Einstein might have been, but I think it wouldn't be even close to the real thing. because we're just missing too much.

Like most of the real Einstein is lost. Because we don't have any recordings of it and the people who knew him personally are dead. And that's out. And it's a shame. But this is probably somewhat tangential because who would know the difference anyway, right? Exactly. And am I really myself when I'm teaching? I mean, I'm different than I am with my kids, right?

I mean, what is the whole you? But the core question I have is, you know, is this going to take one of the jobs? You know, people are always worried about AI taking jobs away from people, from actual, human people. And it seems to me that this would be one of the simplest things to take away, right? I mean, my students sit there in class. And I do things differently because I do experiments and I do flipping, you know,

where we solve problems together and I work around with them. But, you know, other professors don't let them use AI. I make them use AI because I think it's going to be a tool. It's like not letting them use, you know, a computer was or a spreadsheet, you know, 20 years ago. it's crippling them. And so, but the question is, you know, if we can have this, you know, these education companies that are basically what universities are. I mean, the third most important thing that universities do, according to Isidore Rabbi, was, you know, to teach. Like it was getting fun or doing research, getting money. We'll talk about that later,

I hope, with modern academia in trouble. But, but yeah, I mean, why learn, you know, just quantum mechanics. Forget about, like, do I get their whole self? Do I get their whole personnel? Of course not. But can I learn quantum mechanics from an AI holographic, you know, I mean, think big. What is it going to look like in 20 years? I mean, do we need professors to teach? Why can't we just do research? Yeah, I think it's definitely going to take over a lot of these education jobs. And I mean, look, I'm sure you're a great professor.

And I mean, I've seen you, you know, your examples in your videos. Like, I've never been in your lecture. But I can kind of extrapolate from that. And I'm pretty sure you do great lecture. Many of my professors were not particularly good. And I'm pretty sure I'd have been better off with even today's AIs just because they would have been more, you know, more personalized in some sense. They would have been more attentive to what problems I am facing rather than just, you know,

writing down their lectures and not caring if anyone even listens. And so this is why I think that there's for real, you know, a lot of those people, who don't really care about teaching and do it because they have to, are going to be replaced with AI. But there is the question, like, how much does this human element matter? And I think there's going to be like a long tail where humans can do things that are irreplaceable for a long time. So I think, like, personally, you don't need to worry about it.

but maybe some people have to worry about. I kind of want to parlay that into a question about what AI can't do in a certain sense or may not be able to do. And I'd love to get your opinion about this. So there's this phenomenon called lock-in. I don't know. Have you ever heard of lock-in? It's basically the first technology.

Oops, sorry. Yeah, right. You mean in computer development? Yeah. So it's in computer development. There's also, you know, there's an example, allegedly, the space shot. For example, the Hubble Deepfield was only taken because of the width of two horses' rear ends. Have you ever heard this described before, Sabina?

No. Okay. We'll see if you think I'm a good teacher after this. Okay. So the original chariots and roads in Rome were governed by the width of the horses that were the form of locomotion, right? A chariot would be pulled by two horses. And that eventually became the width of a standard railgun.

gauge on a train. So the gauge is the separation of rail tracks, right? In the United States, when they were building the space shuttle, the rocket boosters that would launch the space shuttle, the solid rocket boosters back then, were made in Utah, which is in the western U.S., and they had to be transported to Florida where the space shuttle was launched from. To get there, the engineers knew they had to get through three tunnels, three train tunnels. And so the width of the rocket, which determines the specific impulse that it can impart, which determines the maximum altitude of the spatial, which carried the Hubble telescope, can be traced to the width of a horse's butt back in Roman times. Okay. So that's an example of lock-in. The technology,

the Hubble deep field has nothing to do, but it was set by something different. I want to ask you the following question. We have very good LLMs and GPUs. They're married together. It's beautiful. They're doing something very special. Is it, is there a danger in that they're so good at what they do that they're only going to be good at what they do like that, and they're not going to be able to do specific tasks other than language, perhaps, very well, because they were originally designed for video games. I mean, GPUs were designed for video games, and the neural nets were designed for mimicking human brain activity.

And I've got the brain of one of my arch competitors here. This is someone's brain that we don't like. But are we prisoning ourselves? Because of the success of GPUs plus LLMs, is a good. going to limit our ability to do new physics, for example? Well, so they're actually, I mean, people are producing processes that are particularly well adapted to neuronets. Like, and they're actually called NPUs, I think, in neuro processing units.

And you can push this even further. Like, people try to actually, you know, how the human brain works by actually. by actually using physically changeable connections to encode the weights and so on. So there's a whole industry behind this. So I don't directly see this worry, but I do see the worry, this locked-in issue, is often a problem. If you look at the basis, it's an economic problem. Because if you want to transport this thing, you're not going to rebuild the entire railroad system

and all the tunnels. Like, you could do it, but that'd be insane. And we have the same problem, like, for example, why is it so hard to get off fossil fuels? Yes. Why is it because our entire economies build around it. Like, this is a locked-in problem. And you could say, well, it would have been more sensible to, you know, build battery-powered cars much earlier. But we didn't.

So we're stuck with it. And now we will have to invest a lot of money to get out of it and onto a bigger maximum, basically. So, yeah, these problems everywhere. So if you look at the current large language models, they're facing a big problem there, which is that they've invested like hundreds of millions of dollars into building these models. So that makes it very hard to say, okay, let's forget about a large language. models, we need to do something entirely different because then they have to start all over. So they have a huge bias on trying to tweak these models a little bit to make them a little bit

better, you know, to get to a local maximum, if you see the visual I have in mind, whereas you might have a much higher maximum after you've crossed the valley and put in a lot of money. But I think that this is going to be a problem for specific companies. So what I think is going to happen is you're going to have new companies which develop AIs based on a new system. You know, we can speculate about exactly what this would be. And then they're going to overtake the large language models very, very quickly at some point. We're not at this point, and it might be a few years off, but I think that's what's going to happen.

So it's quite interesting. So no one really knows what GPT5 is actually going to be like. So I've just heard people say it's going to be different than the other ones. So no one really knows what they're working on. And so also if you look at companies like meta or Google DeepMind, they're actually doing a lot of different things. So I think they, you know, their risk averse. So I think they see it coming that large language models aren't going to be, to get us all the way to AGI.

So we need something else. And they're trying to figure out what that something else is. You made a wonderful video recently about quantum computing. And I wonder if I could, you know, provoke you to respond to this kind of snide comment that I like to make about quantum computing, that quantum computers are great at simulating quantum computers and breaking encryption. But essentially those two use cases, which were predicted by Feynman and others back in the 80s, between those two narrow use cases, do you think the promises of quantum supremacy and other things are overblown? Or do you think that there's still potential that justifies the hype and that

they're not, as I call them, two-trick ponies? I think it's always useful to look at who puts money into these things. And I think it's quite telling that there are lots of banks and similar insurance companies and so on interested in that kind of stuff. And why is that? Because one of the applications that they want to put on a quantum computer are some of those optimization problems that you could use for asset management, where they say, well, we could actually invest our money much better if only we had enough computing power.

these problems are too complicated to solve, therefore we need a quantum computer. And so the thing is that in this trading world, even small computing advantages can make tens of billions of dollars in difference. And I think that is driving a lot of interest. And I've never actually worked myself on quantum computing. I just follow this from the periphery. I think it's not totally crazy. There's always the. question like in practice, how much of a difference would it actually make? So no one knows. But I think they see the possibility that you can make a lot of money with it. And this is why they're interested in it because they want to be the first to do it before everybody else

can also do it. But will we actually get there? Well, that's that's a different question. So speaking of things that are in the present rather than in the future, I recently. hosted Eric Weinstein at UCSD talking about his theory of geometric unity. I know you've talked with Kurt Gemungle recently. He did an interview and he did a multi-hour exposition of it. And then, you know, we're speaking in June, but, you know, a couple weeks ago he spoke with Sean Carroll on Pierce Morgan and they had a very heated debate. First of all, what do you make of this, like physics being discussed as debates? I've been asked to debate people like Terrence Howard, the actor who's gotten a lot of fame for things like one times one equals two and other

phenomena. What do you think is, is there a point of debate? I mean, we have peer review and then people like Eric will kind of rail against peer review. And so it sort of sets them in this insulated cage where it's hard for them to get credibility outside of their own, you know, kind of inner circle of cheerleaders. But what do you make of this excitement about theories of everything? We don't even have a grand unified theory, right? We don't have a theory that unites the three stronger forces, let alone a theory of everything. What do you make of this quest for the theories of everything? Is this some macho thing that people are concerned with? Or is there something really important about the approach to having multiple approaches like Peter White or Garrett Leasy or Eric Weinstein

or Stephen Wilfr? So I think these are two separate questions. The first one is, what do we make of this debate culture. And the other thing is like, how seriously are we to take theories or everything? So generally, I like this debate idea because personally, I want to see more debate about scientific topics. I think this is something that science has been missing for a long time. You know, like this public presence where people actually argue about science, you know, on a level that hopefully to some extent.

Your summer starts now with Memorial Day deals at the Home Depot. It's time to fire up summer cookouts with the next grill, four-burner gas grill, on special buy for only $199. And entertain all season with the Hampton Bay West Grove seven-piece outdoor dining set for only $499. This Memorial Day get low prices guaranteed at the Home Depot. While supplies last, price-in-valid May 14th or May 27th. U.S. only exclusions apply.

See Home Depot.com slash price match for detail. The general person can understand. Like the way that we argue about other things, about sports or politics or whatnot. So science has been missing this for a long time. And I'm very happy to see that this is coming. And, you know, I try to contribute my own little share to that. What I'm not so happy about is the topics that they pick for this. And this brings me to the second part of your questions.

Like, what do I make of all this theory of everything stuff? So, I mean, it is an interesting question. and arguably it attracts some attention. But I think we're overstressing this. Like, personally, I think it would be more worthwhile to talk about other topics every once in a while. Like, so for example, I talk a lot about nuclear fusion, talk a lot about quantum computing. There are other topics in physics and science overall than this theory of everything.

Yeah, it does seem to get a lot of attention. I also get a lot of emails about those, you know, kind of theory. of everything. And I've stopped. I've basically started charging. If people want to ask me questions, fine. You can ask me questions, but I charge, you know, $1,000 an hour. And that usually weeds them out. Or it makes it very profitable. And I send the money to charity. That's my new thing. When somebody has a 10 subscriber YouTube channel, I'll come on it, but they have to donate $1,000 to do research. Speaking of research, Sabina, I want to ask you, you're a theorist. Our mutual friend, Lenny Suskin, He has written a book called The Theoretical Minimum, a series of books about the minimum amount of information a theorist should know to be a physicist.

I want to ask you a different question to your students or people that treat you like a mentor and a professor. What would you say is the experimental minimum? What should be the minimum amount of information that a bright theoretical graduate student? What should she know about the experimental techniques, technology, et cetera, say in your field, my field, etc. You just expect me to answer this question like this. Like, I'm going to spit out a list with 10,000 words, like an AI or something. We actually did have several experimental lessons on, you know, how party accelerators work or something and data analysis and stuff like this.

So, and we did have mandatory, I'm sure you have this too, like mandatory. experimental sessions where we had to work with very outdated experimental equipment. I won the Nobel Prize in 1914. Yeah, exactly. Yeah, exactly that kind of stuff. Yeah, and I do think it's necessary, though I wrecked some of this equipment. You know, I wasn't meant to be an experimentalist. That's why I never invited you to my labs.

You wondered why. The Simon's Observatory would be in pieces if you were ever. No, we do that ourselves. Let me ask you a different question then. what experiment, if funded, if you were queen of the world as you are of YouTube and many other domains, if you had this kind of ultimate power and you had the power of the purse, the budget, what experiment do you think would be the most decisive experiment, as Carl Popper called it? What experiment would you most like to see done?

Obviously, you know, you've had your criticisms of, you know, the future circular colliders and so forth. But what experiment would you like to do given, you know, a reasonable budget, say, you know, 30% of all the physics budget for the next 10 years. What would you put it on? Okay, you mean like in the foundations of physics, sort of like overall in the entire world? Anything that interests you, yeah. Anything of the, the Hassanfelder machine.

What would it be? Yeah, well, you can't fund like a billion postdocs. Right. Okay, so the thing is like if it was anything, like in any discipline of science, I would have to look into this closely. Like, I'm sure there are like questions that would have a huge societal impact, but I can't pull them out of my head immediately because I've never thought about it. But I have thought about the foundations of physics stuff a lot.

And this is what my first book was about, basically. So what should we do in the foundations of physics? And the answer that I arrived at was that we should focus on experiments that could test predictions that we arrived at based on results. an inconsistency. And so one of the things is quantum gravity, because we need the theory of quantum gravity to resolve an inconsistency. And there are people who are looking into this, but they're like a few small groups. And I'm like, can we just throw a few billions of this and get it done within a few years

rather than a few decades? So this is one thing. The other thing, which is kind of my pep-peep is that people don't take the measurement problem in quantum mechanics seriously enough. And so in this case, the problem is actually not that we don't have data, but I think that they don't analyze the data that they have. So, for example, I think there's a possibility, let me just leave it that vaguely, that quantum computers might ultimately not behave the way that quantum mechanics actually predicts. And so now you're in this weird situation, that quantum computers actually do things that,

people don't understand. And then they say, well, they chalk it up to, well, it's just some complicated thing going on and they give it a name and then they parametrize it. But they don't think about like what actually means. And so I'm afraid that because we don't have the theory for it, they don't even know what to look for in the data. And so this is another thing where I would say, put a few billions into this, like look for what happens to measurement devices.

as they get smaller or to quantum systems as they get larger. Like, where is the Heisenberg cut? Right? And how much of this can we measure? And there are so many things that you could try to test. Like, what does it depend on? Does it depend on the temperature? Does it depend on the size of the thing?

Does it depend on the number of particles or the density? Or as Penrose says on the gravitational energy or something like this, these are all questions that no one's looking at. And to me, this just seems crazy. because it's like this is something that we would actually need for the technology development in that area. So this is what I would do if you give me the money. Do you want my bank information now? Well, no, I want to do something very risky.

So, you know, I'm the principal investigator of the Simon's Observatory in Chile. We just got first light, finally on the four telescopes that are down there, three of which are looking for primordial tensor perturbations and the cosmic measurements. or a background B mode polarization. And one of them, the massive six meter diameter telescope can see very fine scale structure in the C&B. But one of the things we're looking for are kind of, you know, departures from the standard model, including things like dark matter and in particular things like axions. Would you say that the, you know, kind of case should be closed on axions given the kind of

opprobrium that you seem to dish and dis out, at least on your ex post? What do you think is the prospect for things like axions or even things like inflation itself? Do you think that we are experimentalists should be spending some of those hundreds of millions, not billions, but on other pursuits? Again, those are two different questions, like axioms and inflation. Let me talk about axioms first. People now use the word axioms to mean all kinds of things. It used to be this very specific thing, you know, the axiore that was introduced by Wurcheck and Weinberg

that came out of the Pachy Quinn thing. and so on. And that was ruled out already in the 1970s, right? And so then people invented other axions. And now they actually use the word axioms or pseudo-axon sometimes. Axion-like particles. Axon-like particles or whatnot for all kinds of scalar fields that have all kinds of couplings.

So that becomes a very vague question in some sense. And I don't even know what exactly you're looking for. Like, is it just scale of fears with a low mass, or is it this particular coupling to the stress energy, the stress energy, the field tensor? Or what is it? Yeah, we get both. Well, the technical way we look for it is through a conversion of the what's called E-mode polarization with the B-mode polarization, which is forbidden to exist unless there's some parity violating or some, you know, turn-simon's-like term that gets added to the electromagnetic-legrand. And so, you know, Turner and others came up with this a long time ago, how it would affect, you know, how you could couple, you know, basically give the photon mass. I mean, they're kind of like massive photons. There are so many of them. They're extremely light, you know, 10 to the minus 26 EV or something like that. The coupling is extremely weak. But they have the right scales and kind of were ruling out, you know, parameter space, both in laboratory settings and in the, you know, could be maybe magically explain, you know, a different kind of whip miracle. They could explain. you know, why these galaxies aren't so cuspy in the small galaxy problem of structure formation.

Postdictions, retradiction. To explain the weird phenomena, we can't explain with dark matter. But I assume you also maybe have some questions about alternatives to dark matter as well, right? Are you still somewhat favorably inclined towards Mond or Mond-like theories? Yeah, you know, I changed my mind about those three times a day. So first of all, maybe let me say, I actually think, for the reason you mentioned that particles with a very small mouse have a good motivation to be dark matter candidates because they have a reasonable way to prevent this galaxy cost of issue.

And I actually read a paper about this with a student. Actually, the student wrote the paper and my name's on it. That's how it really happened. But yeah, so this is why I know a little bit like fuzzy dark matter is like the overarching term because it blurs out this cusp. This is basically the idea. And yeah, I mean, it sounds good to me. Devils in the details.

So even if you find this, how would you know that it actually is a particle? How do you know that you can't also explain it with something else? Like this is this entire what they call the inverse problem. Like, so if you have the data, how do you know which theory actually was the right one to explain the data? This is why in cosmology, like they're always trying to fit together several different things so that they can say, well, there was only this one explanation for which everything fit together.

So now, do you want me to say something about modified gravity or first about inflation? First inflation and then, yeah, yeah, inflation. So inflation has exactly the same problem like this. So we do have some data that you can explain with inflation, but that doesn't prove that inflation was actually the only way to explain this data. So it's very hard to teleport inflation from some other mechanism that could be the speed of light limit, as I'm sure you're familiar with. There are people who say we can do this faster than light cosmology, you know, where we have some funky fields that can do weird things. And I think like on a mathematical level, they would say they would both the same observational results.

And so physicists default on inflation, I think, just because that was the first idea. But I think, like, on a logical level, as philosophers would say, the theory is underdetermined. We just can't tell them apart. And I think this is something to keep in mind. So I'm not against inflation. I'm a little bit against all this inflation potential building, you know, where everyone has, like, five different potentials. And then they try to squeeze some predictions out of it. And I'm like, what kind of prediction is this?

It's like you just made up the put. Ambition comes in all shapes and sizes. At First Citizens Bank, we roll with your goals because we're built for what you're building. Fit for your ambition for Citizens Bank. The potential, right, you might as well make up the prediction. So that doesn't make any sense to me. So that's my perspective on inflation. Okay, now modified gravity.

Modified gravity. Yeah. So I've always thought of Mondt as somebody. kind of phenomenal technological effective model. So it clearly captures some things that we observe. And so now the question is like, it undoubtedly in my mind captures some truth. But the question is, like, is this indicative of some underlying modified gravity that's different from general a relativity, or is it actually an effect that we might be able to derive from dark matter if we knew how?

So this is why every once in a while, you know, there are people who say that actually the effects that we attribute to modified gravity, they're just statistical artifacts, they come from the data analysis or something else that went wrong. So, yeah, so that I think is the difficult question, like, what do you make out of it? And then, of course, there are people who always like, a Mont has been ruled out because of, I don't know, the bullet cluster. Bullet cluster. Yeah, yeah, the bullet cluster.

So, and yeah, that's true. But the bullet cluster also ruled out dark matter, right? So, I mean, if you look at the details, this is just generally an observation that's difficult to explain one way or the other. And there are a couple of other such things like, you know, every once in a while someone claims, we've seen a galaxy that doesn't have dark matter or there's a galaxy that's only made of dark matter. And then you look again at this like five years later, they're like, oh, sorry, this observation went away. It was something else.

Or, you know, it was something without data. So I kind of stopped paying attention to it because, you know, there's a lot of headlines that just disappeared. You never hear of them again. So basically, unfortunately, I have to say that the situation is still inconclusive. So we still haven't been able to rule out. modified gravity, we still haven't been able to rule out dark matter. Where are we going? Yeah, right.

What can we do? Speaking of ruling out things that we, physicists, can't even explain in the first place, dark energy and particularly a cosmological constant. I have tomorrow video with Kyle Dawson, who was the past spokesperson of the DESE experiment, and still a major leader of it, and he was here at UCSD to give a colloquium. We talked about, you know, Lambda CDM, you know, possibly being, you know, in mortal danger, if not killed at the 4.2 Sigma 1 and 36,000, you know, chance of being wrong by fluke. What do you make of these measurements that seem to kill the cosmological constant? Your friend Einstein might have been wrong, right, when he said it was a blunder in the first place.

Right. So I think maybe people need a little bit of context. So what they say they found is that. Dark energy is actually not constant, but it's getting weaker over time. And there's a discrepancy between it being constant and it getting weaker. That is at 4.2 sigma, depending on how you look at the data. So, you know, I'm not sure we need to take these sigmas too seriously. So the issue is like there are a lot of tensions in cosmology, some of which could be resolved.

by saying that dark energy actually isn't constant, but it changes with time. Unfortunately, that's a different time dependence. Like, from a theoretical perspective, that is very confusing. There are also people who have criticized the DESE paper who have said, like, it's actually in conflict with some other data. And the issue with the data analysis, like, this is not something that someone like me can reasonably check. You just look at the data and they say, oh, we have this data and we did this analysis. And this is the result.

And then you're like, okay, if they say so, right, what are you going to do about it? You know, I can't spend several months trying to reproduce this plot. So basically, what I do is I sit around and wait for other people to do it, people who know what they're talking about. So I'm not sure that I'm making much sense. I guess what I'm trying to say is that I think this isn't as high a confidence level as they make it sound. Like 4.2 Sigma sounds very convincing, but to me it doesn't feel that high. Well, yeah, I mean, feeling, you know, and data. I guess the question, you know, is it's odd that we seem to know when many, many more Sigma that there's something like dark energy.

And what's under, you know, kind of attack now is the equation of state, you know, this. additional term that generates the evolution of the equation of state, it can be measured in many ways. The BAO is just one of them. And the fact that there's these tensions, and then there's tensions within the tension. So I've said that, you know, what we really need is a good psychoanalyst. We need a therapist for the field. That's where the billion-dollar Sibina-Hassenfeldar experiment should go. We should have a team of, you know, psychotherapists. But in reality, I think it's, I think the tensions with intentions. In other words, there's a tension between the Hubble constant at early times and late times. And then the different data that see that as well, there'll be a tension between the Lambda being a

constant or dark energy being a constant or not. So how does one navigate here? I mean, which tension should we take the most seriously? Well, you just take all the data and you feed it into an AI and then you see what comes out. I've actually been seriously thinking about this, but if you think about it for a little while, you see a big problem with it. which might actually be the problem, first of all, it's not actually a lot of data. And AI does need,

at least the current AIs, they need a lot of data for training. And you just, in cosmology, you always think like it's a lot of data, like, because there are lots of galaxies out there, but then you look at what they actually measure

and it's like three pixels per galaxy or something, right? And the sample is like 150. So, you know, it's like for training in AI, it's not, to go by. The other issue is, and other people have made this point much more eloquently than I possibly can, is that this data is all, well, first of all, you need to exactly understand your

instrument, because that always has some sort of sampling bias. It has an instrument bias and so on. Then the data, you know, it's got to be evaluated with knowing what the instrument actually can do. Then it's got to be cleaned. It's got to be interpreted. And then often it's, you know, there's some foregarten. which you're familiar with.

And then often the data is already interpreted with the model in mind. So, you know, for example, Redshift is one of those things that they try to attach to the data based on the model that predict the universe expands with this particular redshift. And so if you only have access to like to the final data, that becomes a mess. Like, because you have no idea really what was actually already what already went in. And there are some people who say that, I've talked about this in some more details, that this might actually be part of the reason for why we have these tensions. So let me see if I get this example together correctly.

So there is the cosmological principle, right? The cosmological principle says that if you average the universe on sufficiently large scales, then matter should be equally distributed. So you can ask the question, like, what does sufficiently large scales mean? And you can calculate this in the Lambda CDM model. Now, some people use this to calculate what direction we're moving into, because, you know, like the CMB dipole or you can do the same thing with quasars and so on. Because if you want to know what this direction is, you evaluate this.

relative to the background, which, but now here's the thing, you have calculated by using Lambda CDM. So it becomes this, but you've already put in the model in calculating what, you see what I mean? And so now the issue is that a lot of people who do the analysis, like which model fits the data best, just use the data that has already been massaged by using the model. So this is all, this is a complete mess. Data and astrophysics generally complete mess. You don't want to touch it unless you really, really have to. Right.

I did an interview with David Wiltshire of the timescape cosmology. And that's hopefully out by the time this video is out. But the kind of concept that we live in a void perhaps and that these are mimicking what you'd see if you lived in a universe with a cosmological constant or without a cosmological constant at all or just, you know, some very small or maybe negligible amount of dark energy. Okay, Sabina, I know it's getting late there, but we have a million questions from the audience. I have to ask some of them. Okay, first one, Sabina, this comes from Bobby Cosmic.

Dear Professor, I would like to hear both of your thoughts on the apparent decline of academia in the United States. I've been hearing a lot of negative talk about research and universities. I've been hearing that publication quality and academic rigor are not what they used to be. So how do you respond to Bobby Cosmic? So I have a particular insight in what's going on in American University, in particular, recently. I haven't been in the United States for a long time. Of course, I talk to Americans, like our host, Brian. So first of all, I'd say that, like, the decline of academic research is not specific to the United States.

Certainly not. We see the same in Europe and in other countries that I hear of. It's worse in some countries than others, but that's a different story. I think it's something to do with the overall organization of the scientific system, the way that people are rewarded. We previously talked about the reward function, right, your happiness bonus. So scientists who work in academia, they get rewarded for doing certain things, one of which is publishing papers because then they can apply for more money to get more grants

and it just keeps the cycle going. So they continue to be employed. And this gives them very weird incentives. It makes it very appealing to work on something that's low risk so that you have a high chance for it to get funded, so that other people need to approve of it. And so there's a lot of bias in this system. And they're basically, it's basically completely completely,

it's a completely closed system because, well, first of all, there's peer review, but also if you look at the grant cycle, right, the people who sit on the review panels are the same people you work with. And so there's very little external oversight. And that's bad. And I think it's been going on for much too long. People have complained about these problems, like for 30 years or something, no one's doing anything about it. And they'll come a point where it just falls apart. Like, where everyone can see, you know, people like, this is like a complete waste of money. And I think that actually in the foundations of the physics, we're pretty close to reaching that point. Yeah, that was a question by Doc Dealgood.

Is there any progress in the foundations of quantum mechanics? Well, yes, I'd say actually for two reasons. One reason is on the experimental side, there's basically progress every day. Like, this is really amazing. This is why I talk so much about quantum technology. because it makes new things possible. We were already talking about the idea of testing quantum gravity in the laboratory. This idea only becomes possible because of advances in quantum tech.

And this is like real progress. Like 10 years ago, people weren't even talking about it. Or maybe they just began talking about it. But you get the idea. And this is actually driven by the ability to put bigger objects in the quantum superposition. better detectors, they're pushing all the quantum limits. So there's really something happening. On the theoretical side, this has had the effect that there's a revived interest in quantum foundations.

What's still missing a little bit is people who try to push beyond quantum mechanics, is what I've called physics beyond quantum mechanics, like physics beyond the standard model. I think that's what we need. There are a few theories of models of that type. For example, Penrose's is one, and generally collapse models, but very few. And I think we need more of them, and then we need to get experimentalists to actually test them because I think there's something to find, is what I already talked about earlier. Yeah, great.

Okay. Nemo's YouTube asked a little bit of an obnoxious question. Did you really receive a letter from a particle physicist that you made a video about? It's a real email. And so maybe let me say one thing. And it's what he's talking about first of all. You made a very popular video about six months ago. Yeah, yeah.

It was a surprise to me because I thought it was kind of a niche thing. And I wanted to get this out. This has been sitting in my inbox for a long time. It was an email, a very honest email. in my opinion, that I received in response to a criticism I wrote of the community, the foundations of physics in general, but particular about particle physicists. I said this. So I know the person who claimed to be sending the email personally.

I've met them several times. And my husband also knows him. And so I showed this to my husband, like already when it arrived, We've talked about this many times. And we're both pretty convinced it's actually the real person. And I also look them up and they still work at this place. But they're pretty close to retirement. And, you know, I'm like living that live.

Like, what's the point of pissing on their leg? No, yeah, it doesn't really make it. Did you agree with their perspective? I mean, in terms of the challenge of the field. and ignoring, you know, I trust you implicitly, you know, that you received it and that your feelings are about it and your assessment of it was valid. But did you believe that they had, you know, legitimate gripes about it? And if so, what could be done about? I mean, is there anything practical about their suggestions or were they just complaining?

Well, there's a certain irony, actually. So I took out the details from what I read. But actually, the, so the person was going on about there, there's, you know, certain models that are all crap and it's just refurbished something. I think this was the phrase that they used. I didn't know this because that was a fear that, you know, I don't myself work on that. It would never ever care to me. So I learned something.

I was, you know, I was like, okay, it's the same thing. As I see in my own field. And of course, I've gotten many other emails from other people, like in completely different disciplines, like everything from medicine to computer science, where people say, we see the same thing. in my discipline with this topic or this other topic. But the issue is that those people, they don't want to go on record. So this is like a general issue.

Yeah. And I mean, this is the same problem that I already talked about earlier. So people just have an incentive to make up these things that they can continue getting grant money, whether or not this actually makes any sense. I mean, it makes sense to them personally because it feeds their family. But from a scientific research perspective, it doesn't make any sense. I mean, I think an experimental, well, it's hard to, we should break down, you know, what they, some other time, perhaps when we're together in London next in the fall.

But the notion of, you know, kind of the self-interested, you know, pay for. Own it all. Pay off your home. Travel for life. Drive a Ferrari. In celebration of the world premiere of the Monopoly, big board buck slot machine by Aristocrat Gaming, Yamava Resort and Casino at San Manuel is giving one person. a $1.6 million dream package.

The biggest prize in Yamava's history. Club Serrano members can earn daily instant prizes and secure a spot in the finale May 29th. Don't pass go and own it all. Only at Yamava, celebrating its 40th anniversary. U.N. Details at yamava.com must be 21-20.

Please gamble responsibly. Monopoly is a trademark of Hasbro. Hasbro is not a sponsor of this promotion. Your health insurance from my family, you know, it's a valid one, but to the extent that scientists don't believe what they're talking about. Like, for example, I don't,

I would say the probability that will detect inflation via B mode polarization is very low because there's a huge amount of parameter space that we'll never be able to access even if inflation took place. In other words, we won't be able to do any kind of legitimate testing or mapping of the inflatant's potential because we simply won't see it, but that doesn't mean that inflation didn't take place. In other words, we can't falsify or approve inflation. But I don't think that it's illegitimate for me and my colleagues to propose to do these

things because if you don't, you know, take the shot, you'll never know. And we may see something, which would not ironically be as valid as important for proving inflation, which I don't think is a sensible concept. But for disproving all these other concepts from Stephen Hawking and Thomas, I talked to Thomas Hurtag last week, you know, from their model, No Boundary, all these models that don't predict Penrose's model, Turok, Steinhardt, Aegis, all these models that don't predict B-modes. In a sense, we're doing the opposite. We're trying to falsify those. But anyway, I don't see it as illegitimate. I know that there are people that do it, as you say, to feed their family.

But I think the vast majority of people, they're very interested in, you know, uncovering some tiny, minute piece of the grand puzzle and fabric of science. But we can leave that at there. I have a technical question for you from Phil Kales. Does Sabina think non-local models could disprove the mind as machine argument? Could biology perhaps participate in non-local? physics in irreducible ways, not replicable by digital machine. Well, dear, what I'm going to say about this.

Well, I mean, a general comment that one could make is that it kind of depends on what you mean by digital machines. So it's a quantum computer digital machine. So I think there's a fair possibility that the brain actually uses quantum effects. And if that's the case, then you might need. a quantum computer to get something similar to what goes on in the human brain. I think it's kind of unlikely, but I think one can't dismiss it out of hand. So I'm not sure I understand the reference to non-locality because I don't know exactly what

the person means by non-locality. So if they just mean non-local the same way that quantum mechanics is non-local, then I think that should answer the question. I think that's right. I think that's what. Okay, just one or two more questions. and then we'll wrap up and go on to a very brief extravaganza. So let me just see if there's anything else from the audience here.

Okay, the last thing from the audience. How concerned are you, Sabina, about what you've called AI Slop? I assume it's quite significant concern in science, that these machines can generate infinite numbers of seemingly legitimate scientific research. What is that going to do for the actual scientific method? It's going to be terrible. And we're only just seeing the beginning of it because it's only just like about now that the models actually become good enough to make reasonably looking plots and produce reasonably, you know, something that if you look at it, you can't immediately tell it was written by an AI or it was produced by an AI. And what's going to happen is that journals are going to get completely overrun by this.

And I think we're already seeing the beginning of it. and it's going to be a huge problem because they'll have to figure out a way to filter out all the AI slop. And we've seen exactly the same problem already happening in other disciplines. Like they're just getting over with AI slop and they have to figure out very quickly how to deal with it. And that's a real risk. What's going to happen in science is that it becomes much more narrow mind. it because they'll have to come up with something quickly and a reasonable thing to do is probably to go by personal credibility.

Like, if you look at the person who actually does the submission, and then you run into the problem that it becomes like a close club or something. And that's bad because I think you need outsiders to having fresh ideas. And that's going to hit especially hard people from particular countries because we know that a lot of the junk actually comes from, you know, somewhere in the Middle East or something like there. There are some countries where they have a lot of those paper mills going on. And I think like the publishers are not remotely prepared for what's to come. Like a thing, for example, of the archive, like that, you know, you and I are used to, they

have a moderation queue, right? So if, you know, something comes in and looks a little bit fishy, then their system flags it. It gets sense. it gets on to a human moderator. But there are only so many of those because they don't have a lot of money. They're all volunteers. And it can take a very long time already now. Like, I have a friend who has two papers in the moderation queue and they've been stuck there for two months. And you can just, I mean, maybe, you know, I shouldn't extrapolate from one data point, right?

But that's a pretty bad example. Like, you'd think, like, if the paper's really crap, they should at least find someone to look at it and just reject it, right? But what's up? They're not doing anything. And now imagine, like, people are going to submit their AI slop to the archive because they went through one way or the other. It's going to be terrible. Yeah. So this is why I'm worried about it.

Yeah, I think that we'll get back to, you know, what they did in the 1800. you know, you'll have Marie Curie will stand up at the Royal Institution or, you know, Max Planck and do a demonstration, you know, this glowing, you know, film or this glowing rock, you know, makes this photographic emulsion exposed. Well, Sabina, what else are you working on? What would you like the listeners to do as a CTA for high CTR? Obviously, your YouTube channel is massive. It's been an inspiration to me. I have what the Germans call Mitfreude. Whenever I see, you tick past another 100K on your way to 10 million, please, soon.

Besides science without the gobbledygook, quiz with it. Where else should people, what should people be excited that's on the horizon from you? So actually, my poor YouTube channel has been doing really badly in the past couple of months now. I've been wondering why. So I'm like wondering. You can't say that because you're talking to somebody who's got the square root of the number of subscribers you have. So what do you mean it's been doing badly? You've gotten, you know, 200,000 plus views on every video.

I'd kill for those numbers. No, no. So, so it went down like by a factor one half. So for just like really suddenly. And I'm a little bit distrust about it because I'm not doing anything differently. Everybody's seeing that. Everybody. I mean, I've looked at channels with 6 million subscribers and they'll get 50,000 views.

I have a friend who's got, you know, literally a 6 million view subscriber channel and he's getting 100,000 views, which is less than 10%. I think it's common. I mean, I'm seeing it, but I'm not, yeah, I was never as big as you, so I don't know. And what's your North Star? I mean, what are you optimizing for? Is it views as, you know, is it something cynical to get more views? You get more AdSense revenue.

I mean, I can't imagine that. So what's your guiding metric, yeah. So, I mean, like the, I've given up on the ad revenue. because they have this issue that you really has to be longer than eight minutes or something. And most of my videos are like seven minutes, 30 seconds or something. So I've just given up on this. No, but I'm, I'm, the reason it bothers me is like I'm wondering, like, am I doing something wrong? Like, should, do I have to improve something?

I mean, of course, we all have fluctuations like ups and downs and I've seen a couple of those. but none of that was that sudden such and so severe for no particular reason. Like, every once in a while, you know, I make a video that upsets a lot of people and, you know, there'll be like on a paper and subscribe or something. And, okay, that's fine with me. But there wasn't anything like it. Like, there was no trigger or anything. Right.

So I have no idea. I think that's happening a lot. I think there's just so much content out there, especially with AI channels. I was always, you know, curious why you don't do. videos like this, a podcast basically format or, or, you know, we're in front of a green screen mostly for your videos, but, you know, maybe it's people find it more authentic when you're, I stopped using a green screen about six months ago. Not that I can give advice to you, but, but people seem to like, you know, sitting at a table and, you know, just talking to a camera,

like your vlog you put up yesterday or a couple days ago, you walking in the woods. It's very vulnerable. It's very authentic. What about more content like that? Yeah, well, I've been thinking about about it, but it's always like, there's a lot of information that I need to compress quickly into my news summaries. And the green screen just works very well for that sort of thing. And we have an established routine that I can't just unroot one day to the next. So, of course, I've been thinking about it. Yeah. So you asked some question, but I'm kind of getting cognitively tired. Yeah, well, maybe we'll finish up it.

But I have seen a lot of people's channels, you know, in severe view decline. I do think it's because of the, you know, the infinite number of choices that we have, the AI slop of talking head videos, you know, that they can make an infinite number of, you know, just search the archive and put up some paper and or search your, I think what they're doing is they're searching your feed and then just making a bunch of, you know, shorter form content and trying to front run, you know, you mentioned something on X and they'll say, oh, she's going to make a video about it. But we don't have to talk inside baseball.

I'll ask you a question. I have a professional question to ask you at least about YouTube, but I can't do it online because it's highly confidential. So we'll stop the interview now. Are you working on another book or just? Oh, right. And now I remember you ask me what I'm working on. Yeah, maybe.

I've tried working on another book, but it didn't progress very far. So it's stuck at something like 5,000 words, which is not a terrible lot. I've actually recently, you may laugh about it, but I've returned to working on a paper about the foundations of quantum mechanics because I feel like this is like my life's work and I need to finish it

and I need to get it out before I die. Well, I can't wait for that to come out. You'll come back on hopefully when that comes out. So quiz with it, I'll put a link to it. I might make a short just where we talked about quiz with it and I'll put it up as an homage to you and everything you do as a kind of an advertisement, because I use it and I find it very helpful. There are very few tools just on a technical level that scrape the YouTube transcript so

flawlessly and so efficiently so I can use it. I'm so happy you say this. You wouldn't believe how much of a headache it was to get this. I know how much of a headache it is because I try to do it myself with my 14-year-old kid and it just doesn't work. So I have them making 3D printed brains instead. But Sabina, we'll see each other in a few days by the time this comes out. I will see each other in London.

I can't wait to see you. Maybe we'll do an in person quick post because we haven't been, well, we were in person at Peter Thiel's house in 2018. It's been a while. When your first book came out. And then before that, it was 2010 in Stockholm where you invited me to give a lecture on a Foundations of Quantum Gravity or Future of Quantum Gravity Conference with Lee. This is 15 years.

Crazy. I know. My wife was pregnant. You were pregnant. think. I don't know how you do it all, but you do it so wonderfully. Sabina Hasenfelder, thank you so much. Thank you. Free will or determinism, quantum computers are classical. These are questions that touch everything around us, from AI development to quantum mechanics itself, to how we live our daily life. I thoroughly enjoyed Sabina's perspective, and I know you'll enjoy this deep dive episode

with Robert Sapolsky, where we explore his research on behavior and biology that led him to conclude that free will is a complete illusion and what that could mean for society, justice, and human responsibility. Don't forget to like, comment, and subscribe. There's a new way to Sweet Green. Meet Raps. Handheld, hearty, and made for life on the move. With bold, chef-crafted flavors, fresh ingredients, and over 40 grams of protein, they're built to satisfy without slowing you down. Try Raps today in the app or at order.com. Available at all participating locations. Security program on spreadsheets, new regulations piling up. and audit dread? It's time for Vanta. Vanta automates security and compliance, brings evidence into

one place, and cuts audit prep by 82%. Less manual work, clear visibility, faster deals, zero chaos. Call it compliance or call it Compliance. Get it? Join the 15,000 companies using Vanta to prove trust. Go to VANTA.com slash calm.