Unzicker exposes how modern physics uses complex math as a crutch to hide its failure to explain the actual mechanics of the universe. He rightly argues that until we understand why constants exist, we are merely accountants of nature rather than its masters.
Install our extension to search inside any video instantly.
When Physics Substituted Math for Mechanism - Dr. Alexander Unzicker, DemystifySci #417Added:
Welcome back to Demystify. I'm Shiloh.
>> I'm Anastasia.
>> This is the podcast where we search for the theories that are going to change the world and one of our friends in that quest, a longtime guest of the podcast.
We have with us again today and that is Dr. Alexander Unzuker from Unzuker's Real Physics. Alexander has written a new book called The Physicist's Guide to the Age of Super Intelligence, in which he does a really excellent job of cataloging the remaining open questions in physics and what the path forward is for actually trying to figure them out now that we have these incredible tools that can go through the entire stack of human knowledge and pull out things that people have missed. In our conversation today, we go through the sorts of things that Alexander thinks are the biggest pieces that are missing from our understanding of physics. And what it all comes down to is the constants of nature. Physics equations are filled with these components that are necessary to make the math work out that have units associated with them sometimes that don't have units associated with them other times. They're just numbers that are necessary to make the math fit reality. And there's tons of these constants. It's not just one or two.
There's tables and tables of them that are applied for everything from the behavior of springs to the behavior of heat to gravitation to light. They're everywhere. And the problem with them is that there's no physical theory for any of them that can predict their value from first principles. And if physics is about understanding physical reality, then it's pretty weird that we have all of these numbers floating around that are necessary for us to be able to do calculations, but that don't really have any kind of true material significance.
And the fact that they don't have a true material significance perhaps points to a shared overlap of problems that we are facing with AI, with corporations, with our government, which is alignment. That is, how do we bring the project as it stands into alignment with its mission as it was conceived initially? We've spent years obsessed with the alignment problem in physics, which is how do we actually bring a material basis back to our fundamental understanding of physical reality? We have a book that is actually on pre-sale right now where we outline our ideas about that approach and in many ways it overlaps with Alexander's approach that we'll be discussing today. If you want to check out that book as soon as it drops you can follow a link in the description. It is called paradox loss, the material principles of natural philosophy, where, as you guessed it, we believe that the solution to the paradoxical nature we are fed from an early age all the way through college and university comes from realigning the project of physics with a search for a material basis of that knowledge. Our conversation with Alexander touches on a lot of the stuff that we talk about in the book and intersects nicely with the book that he has written, which is trying to figure out, okay, we have all of these incredibly powerful tools.
They're doing all of these things in the world. Are they going to be able to fill in the holes at the foundation of all of our physical theories, or is this going to be something that's more of a collaborative process? And in a lot of the details, we don't exactly agree with Alexander's picture of the future, but that's what makes for an interesting conversation. We get to lay our ideas out on the table. He lays his ideas out on the table. We can go back and forth.
And the reality of it is for the listener that the truth is somewhere in between. And so we are excited to hear what you have to say about the ideas presented here.
>> A huge thank you to our patrons who make this project possible. We spend all of our time studying nature and bringing you these conversations, writing books, doing all of the things that we do in no small part because people have volunteered to give a small piece of their paycheck to this project. And it really is a game of large numbers. For as little as 75 cents a week, you can join our Patreon community. Make sure we never have to cave in to sponsors, advertisers. really pay attention to anything other than bringing you the best conversations and materials that we can muster. You could also support the project by picking up a copy of our latest music project, which is from our band, Secretary of Nature. You can go ahead and stream it pretty much anywhere, but if you like the music, you might consider picking up a vinyl copy of it because that's where the material reality meets the abstract. That is at the heart of everything we do here at Demystified Sai is trying to map a bridge between material reality and the wildly complex abstract reality that we all dwell in. And a huge thank you to all of you who have actually bought this record so far. It's been a lifelong dream of mine to press a vinyl record and it makes me so happy to actually be able to get these out into the world in a form especially as the music industry literally evaporated under my feet as I was growing up. So thank you for buying it. is making sure that we have the funds to actually do our next record, which we are hard at work at, and it is going to be a completely different project that I'm super excited to share with you next year sometime.
>> One last thing before we go, if you enjoy this podcast and you want to help us out, the easiest way that you can do that is you can just hit subscribe. The numbers really matter when we're recruiting guests to come on the podcast because they want to know that, you know, this is a big place where things are happening and we're important to people that are just as important as they are. And the only way that they know that is if we have a big number after our name on YouTube. So go ahead and subscribe. It helps us an absolute ton.
>> Enjoy the conversation with Alexander and check out a little piece of our music as we lead into it. We'll see you next time.
So I was trying really hard at some point to get an AI to draw me a spherical object. It really blew my mind at some point when I realized it had no idea what a sphere was. Like it had lots of pictures of spheres, but it didn't understand that the sphere had a surface. Like for me, in my mind, a sphere comes from my experience playing like basketball or or spaceball or something where you actually handle spheres all the time. And you understand that they have a surface and you know that surface that can interact with the environment and you you have you like have all of these properties that define a sphere for it. It was just kept trying to cobble together pictures of spheres that it had encountered, but those pictures didn't actually reflect what a sphere was. There's one point you're right here because what the models still lack evidently is this everyday experiment uh experience of little kids playing with a soccer ball and so on and football and so how this intuitive understanding how in objects move in three dimensions and what a liquid is and all this stuff. Fair enough.
Correct. That's why by the way why uh Leon the the chief officer of MEA left now and made this research in in that very direct direction. So okay on the other hand I would really challenge if you say or if you mention really understanding or true idea or I mean just let's be experimentalists and uh what jobs can they do in there? I mean there are uh legions of of researchers out there now building every day new benchmarks what they can solve or or not solve. I mean there are kind of interesting benchmarks out there like like the arc arc agi3 or something I heard a couple of days ago it's interesting what they can do and and then we learn from it. But I would really challenge this um statement really understanding because we ourselves do not know what what real understanding in that very sense is.
>> We build an internal connectivity of all of these different concepts to each other that includes >> so does the computer. there's some fundamental way in which it does not build out an internal un an internal worldview that's consistently applied.
Like how about we turn this back towards physics because your book has physics in the title. And so how do you how do you think all of this that we've been discussing so far is going to play out in terms of the future >> of humans understanding the physical basis of their world.
Yeah, I mean we we can directly uh continue here with the will to survive, right? So um well if you assume as a working hypothesis there will be hu super intelligent machines and they want to will develop a will and want to survive and then the question comes up of alignment. Okay, let's leave out that I don't think alignment will work.
However, um the goals of a super intelligence, what what will that be?
Okay, what what's a super intelligence?
Why would it be interested in physics?
Because it would be interested in its own survival, how all came about. Of course, that includes also all the history of humankind, of evolution, of biology.
And uh now we are entering chemistry and also entering the basic laws of physics because chemistry is only possible because we have the unexplained number 137 which puts an end to the periodic table. You can't have elements with more than um 137 protons because otherwise the electron would be would be uh velocity would be superior to the speed of light in the at the innermost orbit.
And so yeah um in short super intelligence will be interested in the fundamental laws of nature because it's in its own existential interest to explore that because it will want to know why is it 137 why is it the electron proton uh mass ratio why is it 836 because if it would that number would A5 there wouldn't be electro there wouldn't be tubes there wouldn't be technically even in this respect no no no development of of digital intelligence or digital life so it's an existential question physics decides whether these or and how these uh beings came to life >> what what do you what do you mean by physics by the way I know this is a stupid question >> a physics elementary physics question is why The fine structure constant has the value of 137.059 and so on which is I would phrase it um the ratio of the innermost electron in an hydrogen atom to the speed of light.
Two very fundamental uh velocities if you want in nature the two most important velocity. What's their ratio?
It's a pure number. And as I said, if that ratio would not be 137 if it would be just five. So we wouldn't have elements uh heavier than atomic number five. So no carbon exists, no complex molecules, no life, no way. Very boring universe.
>> Is but isn't that sort of the same as asking the question, why is pi 3.1415?
>> No, no, no, no, no, no. That's that's pure logic.
>> It's geometry, right?
>> It's geometry. It's it's >> But but isn't the atom just probably an architectural >> geometric a whole bunch of numbers, interesting numbers, oilers numbers or E or whatever in mathematics you have. I mean these are if you want abstract it's a nice playground an intellectual playground but not necessarily everything is relevant for um for physics or survival I mean there is a pretty good explanation why pi is the number it's just the number if you live in two dimensions you can define it it's it's pure >> what's the what's the explanation because it seems like it's just a geometric truth like it's >> a geometric fact.
>> You defi, right? So you define a circle as being this thing that happens when you extend this line and then rotate it about its center.
>> Yeah.
>> And the pi is kind of an acceleration ultimately. I mean and so you know you can you can define why that relationship holds across all circles because you defined a circle essentially using that same set. But why would it be any different for one of these constants of nature? I mean, they're going to reflect some architectural aspect of the physical environment, right? I mean, what's the difference between what I'm talking about with pure geometry and physics? And I and I would say it's the material reality of the situation, right? The the atom is a material process that we can observe and we can inter we can deduce from our experimentation with it >> and then we observe that there are these constant ratios that fall out of those observations and our job is to reconstruct the material reality right >> yeah I mean that that that's related to what I'm always saying we the discussion we had last time about constants of nature and and our job is to explain these numbers. I mean I would like that we explain numbers as the fine structure constant as the proton electron mass ratio in terms of pi or e or whatever that would be a fantastic explanation but it's our damn job as theoretical physicists to do that or even try that you know um so um I don't we definitely that's an unsolved problem yeah and on the other hand in math there is nothing unsolved olved regarding pi. I mean once the the Greeks discovered this is a ratio and then until the proof that pi is transcendental in 1882 uh all is done. There's no remaining riddle why exist pi or why is it the way it is? It's it's it's a it's a >> it's a descriptor of a circle, >> but it's it's but but defined structure constant is an unsolved problem of physics evidently.
>> But it seems to me like if if you were to explain if you were to explain the fine structure constant in terms of pi all you would have done was to deconvolve it into a series of circles at the end of the day. It's like taking a Beethoven symphony and doing a foyer transform on it and breaking it down into a series of circles essentially a bunch of angular uh I mean you can call a sine wave a circle. It's a description of a circle in some sense, right? So you could you can break things down into their components. It's much like the task you were talking about tiling the interior of the cylinder earlier, right? You're just using like unknown geometry to approximate a more complex geometry.
>> Yeah. I mean if you once you have a concrete formula in your hands you okay that would be the case but it's not it's not just a matter of of finding some fancy combination of of integers or pies or whatever to come how many people do that um but of course that possible formula would point you to a theory which as a whole body of mathematics presumably would have to justify that, right? Okay, just if if you just think about the formula, this could be just a starting point or a coincidence to bring us to a complete understanding why this number and not another. It could be that there is a mathematical construction like combined pies. It could be also related to the logarithm of the of the size of the universe of derx numbers and then you don't have this link to mathematics. So, it's still undecided.
I'm just saying that >> it's an unsolved problem in physics and I I will be interested not because of a fancy uh love for physics or philosophy or whatever because it's in his own interest of survival. He would like to this intelligence would like to survive would like to understand the universe where it could possibly go if there are other constants find structure constants and so um he would help us presumably in finding these elementary in solving these elementary questions.
>> I guess I'm still trying to understand what a satisfactory explanation would look like because I said if we could break it down into circles would that count for you? And you said no. And I'm like, yeah, that kind of makes sense because, you know, I wouldn't try to explain a Beethoven symphony by doing a foyer transform on it and telling you the, you know, angular velocity of the circles that it deconvolves into. That wouldn't really explain much about the symphony, right? Like what is a what is a satisfactory physics look like to you?
>> Yeah, maybe the Beethoven symphony is not a good analog.
>> Well, maybe the atom is the symphony, right? And you're trying to explain it.
You want to understand you want don't want to understand Beethoven in the sense that why did you compose that symphony in this way and not another that's not because it's beautiful and I liked it that way and period and but the question is um if if you observe these numbers these constants of nature yes there is always the question why this value and not another and that's how the big revolutions in uh physics came about. It was it was uh Weber and and Maxwell in um 1857 who observed that oh wow the speed of light is related to these electric and magnetic constants.
Couldn't there be um a relation to that and that eventually led to Maxwell's electronamics to the verification that light is an electromagnetic wave? Wow.
This was the explosion of our modern civilization. What I want to say is this cur curiosity to explain um uh numbers uh or constants in physics drove us to the the advancement of of modern civilization. And so we are well advised to follow that path and look at all these riddles and try to solve them.
Okay. And that there are a number of a number of things as a homework to do is to find structure constant is just one and another number.
>> What is the form of the solution look like to you? Like what would a satisfactory solution look like? Like like let's let's go with the speed of light. What would a satisfactory explanation for the speed of light be for you? Like what would it what form would that take? Would it take a numeric form or would it take a would it take a material representation of the processes underlying the the waves? Like what would that take for you?
>> Yeah. I mean I think you're asking for a model rightly. So >> not a model just I just want to know what what is a what does an explanation look like in physics to you? like what do you >> I mean I can't I can't tell you what the future explanation would be because I I just don't know. I just can tell you retrospectively what has been a a revolution and >> the discovery the discovery of the of the speed of light and the electric was one of these explanation. I would say this was a fantastic uh explanation which is of course the body of Maxwell's electronamics.
In that way we should try to explain other constants of nature and this is what we what we should achieve. Yeah.
>> Okay. But I I'm I I want to ask if you feel like the unity of electricity and magnetism was done in by finding Yeah. Sorry.
Yeah. Electromagnetism and light was done by finding an explanation for that unity. Or was it or do we only just have a description of them that tells us they are the same phenomenon without an underlying mechanism that clarifies why the phenomenon looks the way that it does? Because I don't think that saying electricity and light are the same thing is an explanation.
>> It's one step. It's one step. Of course it is. Of course it is a revolution. Of course it is a big insight. And as I as I have my theory of scientific progress, you know, um I think there are always these three elements. One, there is the visionary idea. It could be something.
Yeah. Gravity could be the earthly gravity. Gravity of the apple and the moon. Newton. It could have to do something with each other. Then you have the second thing is the mathematical theory very sophisticated very skilled math Newton's theory of gravitation or be it Maxwell's equations you need this mathematical apparatus to justify that and at the very end the third element which is the simplification you have not three constants any longer epsilon not mu not and c for the electric and magnetic and c but you link them in one equation that eliminates one constant.
So you're ending up with two. And this is the big progress. But we're not done here. Of course, you're you're saying, "Oh, I want to have an expl this is not the final explanation." Yeah, because we still have two in this case and even more constants which are left to explain.
>> But I think that there should be I think we are suggesting that perhaps there's a fourth piece of this >> because you say there's a a fourth there's a fourth piece. You gave us three. You gave us visionary insight, mathematical theory, and then simplification. Okay, great.
>> Physical cause has to be included in that. It has to be mapped to something in reality that is doing the thing that gives you the math.
And that's what an ex that's what Shila is trying to drive. Can I just say something like like if we take a like the most well understood constant possible like every kid learns this in high school like Robert Hook's experiment with elast with springs right what if Robert Hook was only studying rubber bands and they were all the same rubber and everything and he did all these experiments and he was like wow it's weird like the harder you pull on these springs uh you have to pull on them harder to get this extension and there's a constant that scales that extension and I'm going to call it K and it has this value and he deduces it. it turns out to be a modulus, but he doesn't know that because he's only studying rubber, this one kind of rubber material and he doesn't realize he he's just like, man, my god, there's just this constant of nature that's present in the extension of these springs, right? And then of course he didn't do that because he tried other materials and he realized they had different constants associated with them and then he realized, oh my god, it's a material property of of the and then people go further. They you know use atomic theories to understand the latises of the material and the you know the bonding architecture and you start to actually come up with a material representation of what that constant means for the first time right >> yeah it's an excellent example let me let me break down that that in in my way so you have a lot of different springs right a lot of different constants of nature or messages of nature all these different uh springs delivering different numbers then you you you understand that it depends on the thickness and things like add and then you you end up in simplifying all this mess of different springs in oh there is one modulus of say the material of iron right so we have simplified you have a huge insight in simplifying same process yeah and then you do the other experiment not with uh iron but with copper and whatever rubber and then you have again a mess of different messages of nature different um different constants of these materials And then you go one step further, one even more profound insight and you trace all these moduli back to atomic properties at the very end I would say to Darak's numbers which is electric to um gravitational force is one to the 10 to the 40 right because all the moduli came from uh interactions of electricity at the very end right that's how every material is held together by electric forces and you compare that ultimately to gravity in the spring. So all this all these experiments all the possible thinkable im imaginable springs uh at the atomic level come down to that number 10 to the 40 found by drag which is the relation of the electric and the gravitational force. This is a tremendous simplification a tremendous insight and tremendous progress. Yet we are left with the question why this number 10 to the 40 and not another >> right >> and why not lots of different numbers right so I mean at the bare minimum it seems to point to some head some homogeneity of the material structure that you're studying right like if hook had had found that all materials had the same modulus then he might have concluded that the material was unimportant because all materials were essentially must have had the same substructure which they don't and that's why they have different moduluses, right?
>> But when we look at something like the speed of light, that tells us that has to tell us something really profound about the material basis of the transverse deformations that we're studying, >> which I also think that it's very very the thing that you're talking about is exactly spoton. And there's another piece of it which makes it so much harder, which is that it's very very difficult to identify variations in the speed of light.
>> Mhm. And so our ability to take that >> is the speed of light constant.
>> Yeah. Like our ability to even do that experiment like our we don't have a bunch of different rubber bands. We don't have a bunch of different springs.
We have this thing that's so small and so fast and so difficult to to to modify that >> we're just kind of like I I don't know like is it is it not?
>> You you propose before adding a fourth element.
If I think about Yeah, that's I think it's a little bit maybe included in my mind in that visionary idea of Newton realizing oh the apple and oh the moon are of the same.
>> I don't think so.
>> Yeah. Yeah. Still and and and then if you say okay that has to materialize somehow that's together with the visionary idea is is the non-mathematical stuff. Yeah. But yeah, okay, you could make a a fourth point about that. That's that's fine.
And just to to to complete that, I mean there is a list of constants of of unexplained numbers you should work in.
And some of them you may find numerical coincidences just that as a matter of principles you cannot explain numerically um neither uh the speed of light nor plank's constant that's what I discussed in the mathematical reality because there is just no coincidences available.
So my argument back then and we got in an argument back uh back then was that you would have to justify the very existence of these constants C and H by let's say pure mathematics or whatever and you were leaning more towards the idea that okay also explain constants and explain maybe ether properties and what you need to for transverse sound velocity is is elasticity and density and explain these that way and uh yeah I mean you have a valid point if you demand a material basis and say okay I'm fine with a consistent ether theory would be wonderful already I would still go one step beyond asking okay why that and not another but I mean we can leave that maybe aside for today because actually we were wandering off into physics without AI.
>> Question is also if would be AI interested and I still think yeah it would be would be certainly existential interest of >> well they'll be really good at >> smart beings and I will be a smart being so it will have the same maybe emotional attachment to these discussions or it's it's a matter of of of life and death at the very end. I mean it will be very interested in the quantitative relationships but that's why I think we're pressing this. It's like if physics isn't interested in the material basis of these numbers then who will be like that has to be the the mandate of the physicist at the end of the day because there's no other science to do it otherwise like and and the machines will be very good at crunching the numbers but if they are completely divorced from the material reality of the world and and unembodied how will they ever make physical progress >> and and furthermore furthermore for is a study of physical constants like something like the fine structure constant. It's not just a number that like pops out of nothing. It's that when they take hydrogen and they cause it to emit its red line, they can apply a magnetic field and they can split that and the distance by which it splits give them gives them some information about how the hydrogen is interacting with the magnetic field.
>> Yeah. That's historically why it's called fine structure constant. Yeah.
Right. And so I'm like, that's not that's not magic. That's telling you that there's something physical about the hydrogen atom and its the way that its surface is behaving in response to the applied magnetic field that if you could understand a what magnets are and you could understand b what the surface of the atom is and what it's doing then you'd have an explanation that comes from these sorts of physical first principles that explains to you what the constant means and why it shows up because you're you're right that it's some kind of fundament mental feature of of the universe that appears to show up over and over again. But I don't think it's possible to arrive to a an explanation for it as long as we remain eternally in the realm of mathematics because at the end of the day it's the same thing as like okay well why is pi pi >> well it's because of how we construct the circle and the fine structure constant is because of how the hydrogen is constructed. Yeah, I I go with you along most of the way. I'm just saying that um yeah, of course you need a valid model and and you need um and it's a very good thing to to think about visually and demanding that maybe mechanical model in in order uh all full agreement. I'm just saying that at the end of the day we want to have explanatory power and come up with a quantitative explanation. physics is quantitative because you have these legions of theoretical physicists babbling about the unification of quantum gravity and quantum this way and that way and let's philosophically talk about concepts but just give me a calculation of dur's number then you have a theory of quantum gravity not before I mean >> but you can only arrive at the calculation of dur's number if you have something to calculate like >> agree but if you don't have if you don't calculate at the end of the day of the number, you don't have a theory of quantum gravity yet.
>> I I agree. And I guess what I'm I think that what I'm trying to say is that you're like, okay, >> a theory of quantum gravity would be one that could account for the this 10 to the 40 disparity between electricity and gravity. Okay, >> that makes perfect sense. And in order to calculate it, you would have to start from some place where you say, okay, well, this is roughly speaking. Our physical model, these are the physical parameters of the model. They work for gravity like this. They work for electricity like this. And when we compare them, we discover that the forces the the influence that they exert upon the world is different by 40 orders of magnitude. And that ability comes from their physical structure, their material properties. Because without that, >> you're just putting math on top of more math on top of more math. And then when you keep digging down, >> you're just going to end up with another constant. Even if you have one constant, like imagine >> we can take all >> and forget I'm the one fighting against all this senseless math of string theory.
>> Yeah, exactly.
>> And I I guess I guess I'm trying to ask you. And I'm like, do you think that it's possible that like is it possible to get a math an equation alone, a constant alone that reveals anything about the world? If that constant does not have some sort of reason for it that is not mathematical like if you go to the constant and the thing that makes a constant is just another equation, then what >> I would yeah I would answer with Einstein. I I can't imagine a physical satisfactory theory that contains a number which the will of the creator could have just chosen in another way.
You don't want this arbitrariness.
You always want explanations.
>> But explanations are never quantitative is the truth. Like quantities can describe the patterns that you're trying to >> explain. If there is something quantitative to explain that then you have to do it. I'm not saying that it's necessarily complete or or satisfactory to 100%. But it's a it's a necessary condition. Maybe not a sufficient one.
>> The insufficiency is the mapping onto material reality. It seems like >> yeah cuz I mean okay so what Alexander is saying is he's like well you get to a point where if you map it on to physical reality then you have to ask this question of okay well why does physical reality look this way and not that way and so it creates this kind of bottomless trajectory where okay so let's say that you propose a physical theory for why the fine structure constant is the way that it is on the basis of some material properties of the hydrogen >> then Alexander is very correctly asking okay well why in the world is that the physical properties of the hydrogen? And then you have an entirely new question to ask that is perhaps even more difficult to we've spent a hundred years trying to figure out what the heck the hydrogen is doing >> in order to produce the same thing about you know the rubber bands elastic modulus.
>> Yeah, of course. And so, >> but you've made a huge huge insight if you realize that it's describing its ability to deform, right? Like that's a huge insight that isn't there with any of this fundamental constant stuff. It's not even on the table as part of the project.
>> Your PhD was on elasticity also, right?
So, we're not diving into that too much.
>> Alexandra is going to let me bring it back to a little bit into my into my waters here. I mean, let's say let's say that we we able to explain that fine structure and whatever genius could it be. Um but there is a danger here you know I mean people here that's what I accuse theoretical physicists you you're fine with these numbers just accepting the numbers you're not trying to understand not trying to calculate these if a super intelligence fails to do so if it's just disinterested in in these matters and let's say a natural incentive would be to explore outer space and go to other planets and whatever.
There might be another intelligence or civilization with a a vastly um more potent technology uh and and our intelligence here on Earth would risk extinction. Why?
Because if you don't understand the the laws of nature, you probably miss a whole lot of interesting technology. I mean think about that little progress of unifying electronamics with optics led us to the modern civilization. Okay. I mean what possibly we can maybe not even imagine what kind of insight what kind of technology we would be able to develop once we have explained defined structure constant once we have explained the numbers once we have explained the gravitational constant and so on. So it there is a vast possible territory of unknown also technology which might be relevant and if we are just dump and and happy with our limited knowledge and going out there in the universe maybe other smarter civilizations are just waiting for this dumb species as their future food you know I mean and this is why I say it's an existential question. It's a it's a matter of survival to understand physics.
>> I I totally agree. But I think that the hinge point, the remarkable insight is recognizing that the constant associated with the spring has to do with how difficult it is to deform its lattice.
Like that's the leap. That's the that's the physics really. The rest of it is just playing with numbers. It's seeing the pattern. It's then that leap of making sense of what the pattern means.
That's the physic. That's the physics really. Otherwise, it's just mathematical acrobatics and numerology.
>> Yeah.
>> Yeah.
>> Now, when I ask the AI, you know, I want you to take this sphere. I want you to transsect it with a plane or something or I want you to stick a pole through the center of the sphere coming out one end and out the other. It doesn't understand that the thing has one side and another. Like the concept that that is a representation of an object that has two sides to it that can be you know transsected by another object that is totally beyond its ability to make sense of because sphere for it is just a collection of images and mathematical equations.
>> Yeah. And like back in the early days of the models when they first started doing visual generation they had a really really hard time being able to put objects next to each other. And so you would be like, "Okay, I want a blue cube next to a red sphere with a yellow triangle or yellow cone sitting on top."
>> Yeah. These are particularly difficult tasks. Yeah. Right.
>> But and I think that that's what Shiloh is trying to point to is he's like that's not a difficult task. A child can do that. Like by the time that the child is two or three, that child can can can differentiate between the shapes. You know the little game where they like take the cubes and they put them into the the the the holes that are shaped appropriately.
>> Not only should it not be difficult, but I don't see any improvement in it.
>> It's definitely gotten better, but I think that you're asking for the next level of it, which is not just this superficial arrangement of them next to each other. It's about true manipulation of the object. And so we multiple times have tried to use AI models to generate figures that are complex >> and figures that start with a simple piece and then like Shiloh says, okay, so you have the sphere now do things to this sphere and modify it and there is not a consistent internal worldview that it appears to have where it knows how these things fit together and it knows okay so the sphere actually extends into the page and there's this imagined place where it lives and in order for me to draw it >> with this cylinder going through the center of it, the cylinder has its own properties and it just it breaks in this consistent way. And that's what Shiloh means where it's like the way in which it breaks has superficially changed since 2023 when the first models came out, but not >> not to the point where I'm like it gets it totally. It gets it. It's totally on the same page. I say probably you yourself you don't have an internal world view of what you're thinking.
You're not looking in your own neurons.
I mean how this internal worldview is looking right.
>> That's true. But if somebody's like, "Hey, can you draw me a picture of a sphere with a with a pole through the center?"
>> I mean, I'm a bad drawer, but I can do it.
>> Kind of a difficult argument if you're talking about this real understanding or internal worldview. We don't really know what this is about, right? So, I kind of want to avoid that philosophic track of because it's a little bit like like this famous Chinese room argument you might have heard of. Oh, it's it's like um translating from one language to another. It's like using a manual and you follow the instructions of the manual and then you're going to translate the text without understanding it in quotes, but you don't know what the quotes means and what real understanding is. So don't let the the the empty words enter our discussion in that in that sense given that >> let me let me clarify the Chinese >> I agree with what what you have said is is is okay is correct but it's you easily people easily get on on that philosophical track when discussing what real understanding means because we we don't really know what our brain does in understand >> okay but hold on um so the Chinese room example is interesting. The Chinese room is basically so you're standing on the outside of the room and there's somebody who's locked in the room with a dictionary that has all of the translational rules that are necessary to get >> Are there these rules? There are no rules.
>> In the in the in the Chinese room experiment, >> there are no rules in translation. Well, no. There's just there's there's some kind in the thought experiment.
>> You're standing outside the door. You're passing notes in one language under the door and somebody on the inside is is using some reference manual in order to be able to translate them and then passes them back to you translated. And the question is, can you tell if the person inside actually speaks Chinese or if they're just looking something up in a reference manual? Right? Like that's the point of the thought experiment.
>> Yeah. But but that's that's that's uh that's misguided because there is no such things uh as rules with which you can translate a text because you would need uh I mean we agree that for a for a good job in translation you need an entire understanding of the language and there are no rules in language. That's Vitkinstein who said what's the meaning of a word? The meaning of a word is is its use in context. So it's just the context that defines language, defines the words. And that's by the way how the attention mechanism works. It defines the meaning of words by the context. So So I'm just I mean I really want to get out of these um I mean you had a valid point before, but I really want to get out of these philosophical arguments of true understanding. all these folks talking about uh I demonstrate that it will never be like this and that without mentioning a single time how a neuron or synapse works. We're not that's not going to work the discussion here. I mean on the other hand I mean you should should also say okay so how do how I am doing all this with my brain which is essentially neurons and synapsis. So have you ever asked yourself where your internal understanding is and in what your real understanding consists of? Probably not.
>> Surely you don't think that our understanding of neurons is complete, right? I mean there's like I mentioned the frequency domain because they don't function as linear processors the same way that we design our computer architecture. There's an incredible network effect in in parallel processing in the frequency domain that is just like mindboggling. Like I would listen to these dudes in my lab in the next lab over talk about this stuff. I mean it just seems like we're centuries away from understanding how neurons actually work. uh you're mentioning these 3D uh objects like spheres and geometry and the 3D imagination and that's that's an essential of course element of intelligence of and I myself was I mean when these models came out the one of the first things I did because uh let them do some of of my my favorite math problems that is you have um a cylinder inscribed in a sphere and then get it get the maximum value of a of a inscribed cylinder, right? You can make it very narrow and long. It's not the optimal for for the volume if you if it becomes very broad neither. So the optimum is in between and uh it struggled to find the solution and uh so I had as a teacher similarly give a little hints uh I don't think the the hate is always worth the same or or things like uh that's not really what inscription means very general advice of that kind and in a certain point it understood it had to apply the theorem of of Pythagora right rectangles I didn't tell anything about rectangles or or a projection which in this case you have to have the the picture of a projection of three dimensions to two or anything like that so he came up with the idea that he had to apply that I mean what else how else would you like to call that if not intelligence so um and of course everybody has seen examples where they struggled, where they made made uh errors and I mean >> well that's a method that humans came up with right like m like breaking things into their infantessimals and you know deconvolving things into flat surfaces to to compare surface areas that's an established method for for solving this sort of problem right >> so this actually uh relates to something so my brother actually works on Google Gemini he works at deep mind >> and so he spends a lot of time training these models and he also uses a lot of AI tools And we had this really interesting conversation recently where he was saying that he felt really liberated by the way that these machines worked because he was able to just send an agent into the codebase to figure something out and to come back and give him information.
>> And then we started talking about what it meant for that solution to be novel.
>> Mhm. And I think that the the point that is unique to humans, which is that humans periodically have these step-wise transitions where they invent something new, something that has not been seen before, something that emerges. Even if you want to say that, okay, so the human mind is approximately an LLM embedding space. It has all these concepts that are close to each other. And so you have this math that you can use.
Periodically, somebody basically finds some new embedding, right? It's not it's not something that somebody's seen before. They put something into thought space that maps to all these other concepts in a high >> dimension.
>> Yeah, I guess so. Right? Because it basically it adds another layer to the model in the sense of okay, now you have this term. So, let's say I don't know force, right? Somebody invented the idea of force at some point. You can point to Aristotle with his idea of, you know, this doctrine of forms or you can point to Newton with F equals MA, but it wasn't there before. And so that moment of insight is something that's very, very hard to track because I was trying to get Aloha to tell me. I was like, okay, so you have this code base and the codebase is supposed to do something, but it's, you know, 10,000 lines long.
You can't hold it in your mind all at once. You send a machine in that has a higher context window and it solves the problem for you. But does it solve the problem for you in a way that you've never seen before? Or is it solving the problem in a way that is the conventional way of solving the problem?
And so if the answer is it solves the problem in a way that is mapped onto the way that we solve these problems.
there's a category of solution and it identifies that this problem fits into this bin and it knows that this this bin belongs together that's one thing but if it's able to be like ah you know actually nobody's ever done this before but I think that you can optimize the architecture of the way that you call this process through memory and do this new trick and this will restructure the entire system in such a way where not only this program works but you can generalize it out to other problems and I have solved a novel problem. That's kind of like a research question and they can't do that. Humans have to be there to do what you've done with getting it to recognize the Pythagorean theorem application to inscribing a cylinder by a sphere because they don't h they they lack the ability to make the jump. And so that's the piece that I think both of us are pointing to where it's like, okay, I've never seen it make a jump like that. M >> okay there is a number number of possibilities I could continue now just maybe maybe maybe briefly here Demi Sasabi showed a long time ago with these computer games of of the 1980s you know that the computer came back with a novel solution unknown to all programmers at the time it's just one example where you step outside you want to step outside the box I think that's basically what you want to do >> wait what was the what was the conditions of this I don't know about this >> no we just learned from visual input.
Yeah, it's one of these one of these examples Habis in his talks likes to mention that it's I don't know how it's called not Space Invaders but something similar. It's like ping pong and you you you ping pong a little ball and every time it hits uh the cake it one piece drops but and the the clue is to get back to the uh yeah it's hard to hard to explain but it it found let's say it found a novel strategy unknown to all the computer scientists working on the problem. So I would yes challenge from the very beginning the concept computers are unable to think outside the box.
there are kind of this is one thing then before yes I mean the issue of of higher dimensions and embeddings is a very interesting one and as a little sideline I want to mention that um we can we can process information because we are able to see the dimensions the two and three dimensions and and that eases a lot if I see a face which is two dimensional information. It's much easier to process than a string of one-dimensional bits, right? So, but our brain is highdimensional because um how how do dimensions distinguish? In two dimension, you have two neighbors. In two dimensions, you have let's say eight next neighbors of a square, right? 9 minus the central square. In three dimensions, you have 26 um next neighbors. Yeah. attached in a cube. So if you think that if you have 5,000 next neighbors as every neuron has because the the neuron which is linked uh with a synapse is a next neighbor in that sense technically your brain is like thinking it depending on the definition like eight or 11 dimensions.
So this is the power of our brain and that's that's the very reason I think why we have these capacities that it's a marvel how we have these insights these associations which at first sight have nothing to do with each other other and then you you see yeah you your uh your brain is smart and intelligence and capable of doing all these things because you have this hardwired highdimensional thinking. Okay. But there is nothing that prevents you to model that high dimensions. Not at all.
Also thinking outside the box.
I mean you're there there's another little aspect of extrapolation.
Interpolation of course it's easier to interpolate knowledge what you already know. So in between it has to be like this. it's kind of hard to go outside and push the knowledge to new frontiers, right? And uh but I don't see I don't see an obstacle as a matter of principle here.
>> I mean, do you use these AIs a lot? Do you use the LMS in your like your daily life?
>> Of course.
>> Yeah. Yeah. Me, too. I I've been using them since the start and I I have never experienced it try something out of the box or throw me a new idea or anything.
I've seen it assemble ideas that I've that I can find in the universe and and cobble them together. But I mean, you must admit that there even if there are exceptions that you could point to that we don't we we I don't understand the example you gave. Maybe Nasia looked it up or something. All right. Yeah. So, I don't understand that example, but I've the fact that I've never experienced that >> and that I experience it constantly even from my cat or from, you know, my my own existence. I mean, I'm constantly coming up with new ways to solve problems. Uh, and even any basic living animal does this in a way that I just don't even experience a a hint of it in the artificial intelligence. And so, you know, you've got to admit that there's a chasm between the way we go about things, right? I mean, do you know uh do you know this uh gentleman uh Elon Baron Holds by any chance?
>> No.
>> He's a psychologist in Florida. We we recently had him on the show and he's he's made this case that the way that humans conduct language is identical to the way that LLMs conduct language and actually it's a very compelling case.
>> Uh but he was very hesitant to conflate the formation of sensical language with this with thought itself because people can have whole conversations. I mean, surely you've probably been to conferences like this where people are just talking and nobody's making any progress. There's no actual like advancement of ideas. People are just using big words that they've heard before. They're just kind of crashing these linguistic uh spaces against one another, right?
>> Yeah. And and that's the chasm I think that we're circling around here is the difference between stringing together things that make sense and pulling on on pieces of information that have already been handed to you versus actually packaging them into new concepts. Right.
>> And I want to add to that because I think that that's exactly right. the people that work inside of these AI companies like Dennis Hassadis. Dennis has a goal which is to convince everyone that this machine is worth the billions and billions of dollars that >> Oh, come on, Alexander. Of course, he's literally >> AI safety companies are funded by the same people that are selling these machines. I mean, it's all part of the giant.
>> I mean, if if if you we can talk about Sam Alman, maybe not the most honest guy on on earth, but >> okay. Personally, I I don't have this impression of Disab I don't I don't think that he's agreement on that. Let's say could you I don't think he's dishonest. You're right. Okay.
>> I don't think he's dishonest.
>> Nonsense. Okay, let's let's just assume for the purpose of discussing that.
>> Well, hold on. Hold on, Alexander. I don't I don't think that the idea is that they're lying. I think that the idea is that it's it's possible to cherrypick data in order to tell a story that organizes the world around you in your favor. And I don't think that's the same as lying. I think that that's good business.
>> You need investors to be clamoring at your door because you have built the machine God. But but Anastasia, do do you have the impression that I am the gullible guy who believes in science who who I mean he isn't aware of of group think of community beliefs and things like that and people living in their own bubble and like particle physicists and cosmologists I accuse all the time. So I'm I would say I'm the skeptic. Yes, I I would challenge every everything, right? But on the other hand, I mean I I don't um I don't challenge every I mean I challenge everything but I don't disbelieve everything as a matter of principle. I mean of course you can have also bad intentions and there is surely a lot of people here with a lot of money and bad intentions and no I I give you that okay but yeah just and I'm not even following the authority of of Altman or Eric Schmidt or Habis whoever authority I'm I'm totally disinterested in here I'm just thinking trying to make my mind up with what I have learned and what I see here and how all this is works that that's a tremendous interesting question right and we all struggle to to to understand this so um and I guess yes there is um a couple of misconceptions here especially many philosophers talking about the stuff without having a clue of neurobiology or never uh thought about how a synapse or neurotransmitter works that's a little bit annoying I think and but this is one thing we already had this the second thing is I mean okay you're pointing out this error and that error and what you didn't understand at that point and and right right these are examples that happened I mean I have seen a lot of errors uh done by these models to error is human okay keep in mind that little kids who are de have developing brains the same brains we are um amazed about how they work as an adult. They make stupid little errors. They they tell you stories. They make all kinds of stupid uh stupid mistakes. This is part of the learning process also. And we as humans we we try to I mean it sometimes it seems to me that we are forcefully looking at all the errors while immediately dismissing every incredible huge progress done by this artificial intelligence. I mean think back. I mean the the calculator is a fantastic thing that multiplies two six-digits numbers in a split of a second. Okay. What is that? I mean, how would how people 500 years ago would have amazed by this technology? No, it's just a calculator.
And so, we had we had dozens of examples and I grew up with um okay, chess is something must be very intelligent. I mean, that that it's the core of my own self-esteem since my father was a chess player. I I loved him for his being ability to to play this incredible difficult game which I couldn't do the same way. Yeah. And a certain point you have to admit, okay, computers could do that. I mean, how do you grapple with that? And then you have the next resort to say, okay, chess chess is just a tactical games. You you just need to do these calculations and well, okay, computers are fast and they're not really smart and then uh they can do this. Okay, but go is the strategic long-term planning. You computers will be ne how many times I have heard that that computers would be never able to play go and it just now we are here you know and every time computers do something incredible and now ch we have has passed the touring test. The touring test is historically the definition of human intelligence. might discuss that.
Okay. But just historically it was and we have passed the touring test and now people coming along. Okay. And seeing all the errors and and all this okay yeah there are still errors right and you people are >> what's what would be really nice I think it would speed things along to being a much more useful assistant is if it could learn from its mistakes with me >> and learn for other people you know I mean I proposed these things are so unbelievably stupid. I like I I I have such a difficult time with them. I they still find them useful for certain things, but you know, I'll I'll be like working in some program and I'll be like, you know, I need to get this synthesizer to talk to this computer module here. And I'll be like, how do I do that? And it'll be like, okay, you just go into this setting in the computer and do this. And I'm like, there is no setting for that in the computer. And it's like, oh yeah, I forgot about that. Right? And and and then I'm like, okay, well, now you know, right? But you know after a couple of days it's forgotten even with me let alone with everybody else who's going to have this same problem right so it doesn't have any like feedback mechanism into the general network that's able to actually teach it so it can't do the thing like a child like I know when I go to try to connect this particular MIDI module to this synthesizer whatever I know how to do it now because I remember having done gone through this hellscape with this computer and figured it out eventually but it won't remember that and anybody else who ever goes to try it won't remember that Right. And there's this this is such a specific instance and the world is almost infinite in its confirmations for which these specific instances could show up and it and and there's no way for the thing to actually learn about these these myriad instances of these mistakes happening all over the planet. That's not really built into the architecture, right? It has to just be trained on some static set of information for which it's going to be used and feed into really bad answers for people like me down down the line.
And so, you know, there might be an iteration of this that's different that actually has this kind of feedback mechanism to it, but it's it hasn't gotten better at that since I've started using it.
>> I use the customer service of Amazon a lot, by the way, and I know where the stupid bot is and I know it from the German telecommunication companies.
Granted, okay, fine. But don't judge the population by its most stupid members, right?
>> But this is all the members of the population. Well, these are the flagship models of, you know, Gemini and Claude and Chachi PT. This isn't like the Chipotle ordering bots. This isn't like the the German telecommunications assistant.
>> These are the most powerful LLMs that mankind can throw up right now.
>> Well, apparently they have other ones behind the scenes, but yeah, the ones the most powerful ones that we have access to. And >> no, I mean, of course, these company bots are not and are very stupid still because they're just not implemented yet. there's as a matter of principle there's no problem to implement also a smart LLM to go into that there are issues of of data protection all things like that but we're talking about the general development of intelligence I think and I think one one important thing I think you guys are overlooking is just I mean we can look at the data points at at the function value but we're going to look at the derivative at the time change you know >> and if you just see the rapid development and Don't argue about that.
From CHP coming out a couple of years ago and now what CL code does, there's a huge difference. I mean we have gone from elementary school level to PhD, postto whatever level in yeah less than humans need for that trajectory, right?
>> Can I give another example? There's one that really occurs to me because you know I spend like half my time doing music. the the AIS uh have you heard any AI music before?
>> Like they're actually they're really good at recycling old music. So they can make a really compelling generic R&B song that sounds as good in my opinion as the really generic R&B stuff that's coming out of Hollywood, right? And it's it's but the funny thing is it's just rehashed R&B music, right? They're they're not going to ever you're never going to like have your AI just wake up one morning and and do something like what Jimmyi Hendris did for the electric guitar. Like they're not going to have that breakthrough insight about whoa what if we like use the amplifier as the instrument instead of the actual guitar itself. You know that's that's a genius level insight that only a human has has ever demonstrated that it's not even a computational thing. It's what what is that? I mean it's it's literally Jimmy looking around at >> Fair enough. Fair enough. I mean I would agree. I mean I mean >> I would say you can tell a computer just composed the 10th symphony of Beethoven and he would do that >> which is a great thing and let's not be too dismissive about that but it's amazing. Beethoven had composed it his 10th symphony. I mean you wouldn't say oh you're just rehashing your nine symphonies. Okay. Yet you have a point that there is another level of creativity and I myself I'm wondering about and it's it's often not just the very technical stuff.
>> No >> like saying deriving Einstein's equations from a from um mineral principle or some things like that. It's things like um Neil's bores inside oh uh the angular momentum of the electron orbiting the the proton could be just the new found newly found constant h of plunk right this flashes of genius uh these associations and and you're right in saying that until now only humans have proven to be able to do that but again I mean I don't see a limitation as a matter of principles because you still have to answer me I mean how what's the fundamental difference between your retina your links to the retina synapses to your brain and all the processing what's the what's the principal difference I I don't see any there so you we can yeah we can go to this discussion of of creativity and how to advance and then that's all interesting impact yet. I would say >> I think the similarities are pretty self-evident. I think you do a wonderful job in your book of laying out the computational similarities between the human brain and the machines. I I think there are a great deal of similarities.
I just personally am fascinated by the striking differences as well because I I sense a lot of anxiety and paranoia about pe you know this thesis that people will be replaced outright by these machines and there is something really special that I notice about us human beings that I I don't see even the slightest spark of in these machines and I think that's our ticket to ride in the future. That's why I'm sort of pursuing the tiny little window in the difference between us because I think that's our true value to the universe. That's my intuition. Anyways, >> pardon the interruption, but I have to ask you for a favor and it has to do with the success of this project in the long term. We started Demystify Sai because we felt like something had gone wrong at the very heart of fundamental physics. Somebody had a long time ago forgotten that material mediators were important to being able to understand how the universe actually works. And over the course of the last five years, we have met with incredible success seeking out the people that are trying to understand the world and bringing their ideas to you. But if this project is to survive and to be able to expand beyond just the realm of physics into all the things we love, we need your help because we don't want to slap advertisements and sponsorships all over this program. We believe that knowledge should be free. Knowledge should not be used as a billboard to help somebody sell their chachki or their online course. And so we need your help to support the podcast to help us escape the fate of advertisements and sponsorships and whatever other things lie down that road. The best way that you can do that is you can come over to patreon.com/demystifiesai and sign up to give us a small donation.
75 cents a week is all it takes. Through the power of large numbers, we will be able to grow this show into something that is far beyond just a podcast and actually help the scientific renaissance come to fruition. And so I'll see you there at patreon.com/deify.
The identification of the unknown problem is something that's really beautiful that humans do. You know, like Jimmyi Hendris looked around and was like, the problem is guitar music is boring. like it hasn't gone anywhere in a hundred years and I want to I want to address that problem and everybody was like I didn't even know that was a problem and he's like yeah it's a problem.
>> Yeah. The genius sees hits the target nobody has seen. Yeah. The talent just hits the target no one else hits but nobody has seen. You're right in that.
Yeah. Yeah. Yeah.
>> And if they start doing that then I'll just you know dig myself a grave I guess.
>> Yeah.
Yeah. I mean that's we want to right.
Yeah. But coming back I I mean I was mentioning this alpha go to give a historical example of how we continuously uh retreat and dismiss what has been achieved by a as just computation or just this and that. But don't forget we're talking about 2016 here. I mean that's andabis back then was talking about the very specific task of a well- definfined game and this and so on. But now we had the the transformer architecture which is based on a 2017 publication by the Google researchers and that additionally changed the world entirely. Right? So we have this conversation like uh thing the we have the touring test we have the translations we have I would say a genuine understanding of what I am saying. Sorry, but um I if I talk to these to these and it's not that they just rehash or come up with the same thing. I give you a little example. I was discussing uh in one of my holidays. I was on the on the island of Malta and I just wanted how much time would it take for erosion to uh to brush that island away? And he said okay 5 million years a very smart reasoning. And then um how is it like I said okay I forgot uh in the entire discussion um if you're talking about these earth ages there has been historical um um um sea surface level have been changing for a whole lot that changes the entire discussion and he said yes yes yes it's true and that by the way what also changes the entire discussion is volcanic activity And I was okay. Didn't think about that.
But it's like if you had a conversation with a smart geologist who not only was able to do all the calculation stuff with the erosion, um, not only uh recognizing that there might be a problem with with the surface level variation, but also additionally come up with an obvious problem in this context, but it wasn't mentioned. Yeah, >> that's wonderful. Unless the geologists have missed something really important about how volcanoes work, which they very well may have, right? And this might lead us into a conversation about physics next, too, because >> you're you're right that that I love it when they point things out like that, but they are also predicating that on some extent body of knowledge that may be deeply flawed as well. And they're not reasoning their way from first principles to that. They're they're they're collecting what people have. It means reasoning from first prince. Be careful about this. This is philosophical talk.
>> Well, hold on. I have I have a way.
>> You don't know what you in your brain are doing. I mean, >> hold on. Hold on.
>> Doctor, it's a philosophy here, bro.
>> Out random words one after another randomly, you know. I mean, that's it.
It's so easily one one could be insulting, but but uh I mean, >> well, hold on, Alexander. Okay, imagine we roll back the clock and we take our technology of transformer architecture and reinforcement learning and we go to the ancient Mayan world and they have their cotices, they have all of their information, they have their metaphysics, they have the system of worlds and then we train the machine on everything that they know.
>> Interesting thought experiment, right?
And then we're like, hey, it's stopped raining and our crops are failing. What do we do?
It does not reason its way to the habber bos process. It does not reason it >> hash. Okay.
>> Yeah. It doesn't reason itself to the habbash process. It doesn't it doesn't figure out like Okay. Well, so there's this thing where if you take plants and you and you breed them and you and you cross them, you can eventually get like more drought hardy plants. It's probably going to be like, okay, so the rain gods are angry at you because you have sinned against them and what you need to do is you need to wage war against your neighbors and collect as many sacrificial uh participants as possible and then take them to the top of the temple and cut their hearts out with the most opulent ceremony that you can possibly have. And then if you take those hearts and you plant them in the fields, then your crops will grow because the rains will come. And there is no passage from that locked frame of thought to the frame of thought that we have today which is about atoms and molecules and and solar radiation and corololis effects and jet streams and everything else without literally somebody sitting at that model and going out into the world and collecting data about how things look and the weather patterns and the the the like the water cycle and climate. It it just it can't happen.
There's a there's a hard wall to what these machines can do that is fundamentally limited by the context in which they are trained. And I don't see them being able to go beyond that because when it comes to understanding, I agree that it's really hard to explain what the hell it means for a human to understand something. And yet, it's one of those things that well, I know it when I see it.
>> Mhm.
>> Because >> Yeah. Great. Yeah. Yeah. Yeah. That's what I mean. I said I said with with a model trained on the epicycles it would be hard to discuss cosmology, but you made it even more to the point. Even more to the point, even harsher. Yeah.
The mayor culture. Go to the temple, sacrifice something. Uh, okay. I totally agree. And that's to be very short.
>> It works last summer.
>> Yeah, it worked last year. It worked for King. Whatever. Yeah, you should try it too.
>> Yeah. Yeah. Yeah. Here we are. Here we are. Yeah. Here we are. So, it's kind of difficult to imagine to have these models without any scientific method whatsoever.
But even then it would be kind of reasonable as you said Shilo um say okay let's try that.
Try to do that. Try to sacrifice and we'll see. um what comes out next year next year again and so on. So once we enter this cycle of this scientific method being applied to sacrifice >> very much patience in that sense okay I admit >> n equals 100 this year >> once we try once we get into this cycle right you have kind of a scientific methodology as a matter of principle he could sooner or later verify that helped or and maybe if it didn't work out so maybe we try either an extra sacrifice or we just do something else or >> Yeah. Yeah. Do you But do you ever see the LL like would the Mayan LLM ever be like, "Hey guys, what if we actually tested out this idea?"
>> That's exactly my point, right? Because the scientific method is this, >> it's a human creation, right? It's a beautiful creation of the human mind.
>> You don't know what that's what I doing every day. I mean, uh, if if I'm I'm I can tell this we'll make a video soon about that. I'm I'm right now I'm analyzing the LIGO data, the gravitational wave data, re um reanalyzing it with a method I think it's better, the the direct correlation method, and he just um if something doesn't work, he just improves it. Okay, let's see. Okay, what's the problem? So, or maybe it's the problem that we had just uh 32 seconds instead of the whole range of 4,000 seconds. and he indep without me even knowing about the problem he comes up with that solution if you want. So I admit in a more narrow sense we als we already have this stepping out of the box and proposing solutions and um well I mean uh it's difficult to predict the future and and I mean we can endlessly discuss about the state-of-the-art right today but just I would just propose to lean a little bit back and see what we have achieved historically the traje trajectory, especially the slope which we're observing red in the moment and compare that to um yeah what um and on the other hand think about what what is it what makes you think you're so special as a human with your particular >> well not me not me I'm not a genius >> I mean yeah at the very end we are I mean we love our species we love ourselves that's okay fine but I mean What's so special? Why shouldn't why shouldn't artificial life, artificial consciousness, artificial intelligence not be possible just on another substrate? What's what's the principal obstacle?
>> I think the biggest problem is that because we have such intelligent people working, let's just use physic. Let's use astrophysics. It's something we're all, you know, fairly familiar with on this show and everything. you've written a book uh regarding Pier's approach to astrophysics. The problem is, you know, if what Pier did by re-examining the sun really in the context of 19th 17th century physics was a weird regressive leap that would be difficult to make in the context of how well the super intelligent quantitative mathematical physicists have talked their way out of all of the contradictions. like they have made a model for the sun that is quantitatively consistent. For the most part, they've been able to use gases and ideal gas laws to explain how in the world you go from having these high frequency photons at the heart of the sun that ends up leaving the photosphere, which isn't even really a destination. It's not really even a surface in their model.
and they have an argument that they've they've solved all the contradictions with it to their own liking, but they may very well be completely wrong still and it may take some, you know, lunatic on the fringe to come along and say, "Well, yeah, you you guess you could say all that stuff, but there's a actually a different way of looking at this that not only would be more simpler, but probably works better." Oh, and and on top of that, there's this other piece which is that when you're doing when you're theorizing about phenomena that you cannot directly see.
>> Mhm. Then you are stuck without actually being able to go and verify to discover if the thing that you have proposed is accurate and you end up with something like a black hole or a neutron star because you can't go and see if the black hole is actually a black hole. You can't go and see if the neutron star is actually a neutron star. You can't even really go to the surface of the sun. We don't have materials that can withstand the temperatures that are there. And so it's not like you can go and throw a rock easily at it and say, "Okay, it splashes. It's got a surface." And so there's this break between the theoretical, which are the things that Shila's made a really excellent point about that they match together into a perfect hole that just is impossible to challenge because every single piece of knowledge is dependent on an antecedent piece of knowledge that >> every time there's an inconsistency, some really highly computationally intelligent person has come up with a very compelling and clever workarounds.
Like I always love to say that, you know, have you do you have any friends who are lawyers?
>> Like lawyers.
>> Oh, you're a lawyer. Nice. Nice. Yeah.
Lawyers can argue their way out of anything, right? They could be totally wrong and they think >> Yeah. Yeah. You are just argue. You job is not to find the truth, but just to Yeah. to develop an argument.
>> That's an entirely interesting. But let's let's leave aside that maybe. Yeah. uh it's it's a big yeah big consequences also for society in this respect but what you said before about the the sun's model actually I mean that that's that's my point I think that uh these models might be able to more objectively discuss and review the evidence we have and of course we have we have u the evidence and the observations and different models and different people draw different conclusions from it. But I think that there is a certain objectivity or unbiased approach by these model. you remember the example when when I mean he was heavily defending the the mainstream um solar model but when I when I showed um just the video of the of the solar flare and he said okay that really looks like it couldn't be a a gas right so I see really a great possibility here again to be more objective more more unbiased, not influenced by any funding or whatever publications or career, all that biases we have in in in real science today, unfortunately. Uh yeah, I I think we can have a new kind of scientists and and of course, okay, you're right that nobody brought that up yet. Okay. But because um you need people to do that. But at this point I mean we in such a privileged moment because we are yet unique a little bit in in that respect of oversight or guidance or or draw pushing the model maybe to the interesting questions because it's not really not really able yet to to to find the interesting things. I wouldn't deny that on the long run. However, at the right now we are in a very good situation to yeah to just let the let the models work on that interesting on these interesting questions and also with with the same the same thing with um ether models of electricity. I mean if you if you go out there to a to a publisher to to a journal they would just dismiss you as cranky if you come up with a ether or a rubber model. Okay, we have heard this 100 years ago. No, it's interesting. So, um I think there's uh there is a big possibility and and to your point Anastasia because we have to rely on the observations of course uh when you want to change electronamics if you want to formulate in a consistent nonlinear way it doesn't help you in this particular for this particular purpose that you have played as a kid with your toys. you need to um make sense of the of the observations. Yeah, the astrophysical observations or or accelerator or whatever. Yeah, that's an abstract thing.
>> Well, you should still probably like consider that the dynamics you observed with your bouncing ball should be maintained in the theory that you propose, right? I mean, >> yeah, >> you can't just have like these zerodimensional thingies doing stuff to each other. Of course, >> nothing works like that on the playground.
>> And and I'm not I don't want to really take a side on whether Leon uh is right with his critique. He has a very princ he has a very principled critique of the of the LMS that they won't be able to do certain things if you don't do that kind of training in the real world as the little kids do. Could be a fair point.
On the other hand, yeah, there are lots of other smart people arguing in another way. I'm not decided on that. I think that's that's at the very end a minor issue. It will be resolved either way. I mean, one day we have the robots and they will they will be training right now. They're building a training center in Munich for robots and they're just learning how the real world works, right? What's the problem?
Well, the I don't necessarily think that it's a problem. I I think that this comes back to the question that you asked, which is what makes humans so special, which is that it seems like our survival is dependent upon being able to identify problems and solve them. Like, if I was to generalize what it is that humans do, it would be that in a way that other animals don't. And other animals don't really seem to innovate.
Humans innovate. Humans look around and they're like, you know, I'm really sick of eating raw meat, but I found this, you know, I found this antelope after a forest fire and it was actually really tasty. And so maybe we can maybe we can figure out how to harness that and then do that at home. And then they figure out fire and then they figure out how to cook food. And once they cook food, all of a sudden they can unlock all of this nutrition. And now that they have nutrition, their brains can grow bigger.
And each next step pushes them >> and they can travel, right? You can store the food now. And >> yeah, like there's tons of there's just all of these things that come down from some kind of observation of something that's happening in the real world, an accidental happen stance. It's like Carrie Mullis who discovered PCR says that he literally had hallucinogenic vision while he was driving a convertible in Malibu where he saw the entire system in his head and then he came back in the laboratory and he tested it and that's how it worked.
>> Yeah. And so there's there's something about just being out in the world and having this problem that you're turning over in your head and you're taking in information about the outside world. And there's some module inside of your brain that just kind of is is is flowing your experience over the surface of the problem that you're trying to solve and eventually it eats it away and then bam you you get it. And so, >> and sometimes you don't even experience that just in your brain. Like I wanted to say I I personally have had this experience a lot lately where you especially writing this book that we're doing. And I've gotten closer and closer to finishing it. I I sometimes get this like sick feeling in the middle of the night. Like I'll just literally my body I'll feel like kind of nauseous and gross and then I'll realize something like I'll have some revelation about how one piece of the book connects to another >> I'll be like oh yeah like we need to make this link >> and then yeah no I just realize it maybe I'll write a note down or something but then I feel I don't feel sick anymore >> like and it's so weird it's so weird it's so uncanny it's like something in my body knew that something was wrong >> but my mind didn't even catch up to it and and then as soon as my mind has has addressed what my body was talking about >> that my body knew about my conceptual existence and then it's very very bizarre. This is why I'm like I don't know if neurons are the whole story or if we understand the I mean there's freaking neurons in my guts too, right?
I mean they were participating somehow.
Like what's going on? This is insane.
>> I mean you're right. That's fantastic.
And there are these stories of Menlay for example he dreamed about the the the table of um chemical elements and things like that. I don't want to belittle by no means the fantastic how how our brain works, how our body works. That's that's really incredible and makes you humble and and that's why yeah of course we are very special and that's why we dominate planet earth at the moment. We we are an incredible species and certainly have uh yeah have this problem solving capacity which is unique to uh which character characterizes as the dominating species on earth.
However, from a biological perspective, we are not I mean we we not in a religious way the the the the the crown of creation. What comes last? I mean nothing special in that sense.
There could be another intelligent being based on biology, maybe based on other amino acids. Um and this is one thing. I mean we are fantastic species. I think we are characterized by our intelligence.
But uh how long I mean we are smart enough to dominate the earth but honestly are we really made for long-term survival in this hostile um environment of the universe? I don't think we can take that for granted and I don't think that as humanity as a whole we are showing much evidence we we are able to address all the problems we're creating right we are not merely intelligent I guess is what this is circling around right I think that we are deeply uncomfortable with inconsistencies in our model of reality in a way that computers are not there's there's something progressive about that in the experience of going from being a child to being a fully grown adult to becoming a theorist or a master artist or anything else where you you really experience something and honestly this is the known versus unknown problem right so solving your computational ability to solve known problems is what an IQ tests right this is computational raw power and both of us the machines and other people besides me are really good at computational power for sure.
But that that discomfort that motivates you to go to work and and look for a new problem that doesn't exist that will set you at ease with your model of reality.
That that seems to be the unknown problem. That seems to be the the the desperate search for the unknown problem which is perhaps going to guarantee our existence into the future.
I think especially because there's a lot of research coming out about the ability of animals to have language. So AI tools have been used by several different groups of researchers to study the vocalizations of whales.
>> Mhm.
>> And what they're finding is that they appear to have we we have no idea what it is yet, but they're able to classify all the sounds.
they're able to identify how different >> research by the way. Yeah, >> there was there was I don't know if you saw this but they just published a really cool paper >> is from Gashbar.
>> Uh this is for Gashbar's group and then others associated with it. Gashpar is um a linguist at UC Berkeley that we've talked to on the show before and he so basically they they found a pod of sperm whales where one of the whales was giving birth and they recorded the vocalizations during the process and they discovered that the vocalizations changed and that they also had this very complex process by which they kept the calf alive because when the calves are born, they they have this organ in their heads that needs to fill in order for them to be able to maintain buoyancy.
And so all of them, so as soon as the calf is born, it's dead weight in the water. And all of these whales come together and they coordinate one at a time to bring the calf up to the surface so that it can breathe. And they're all talking to each other. There's other predators in the water and they're organizing when the predators show up to protect the calf so somebody can keep it stable while it kind of wakes up into the world for the first time. There was somebody else who posted a really funny uh thread about he uh he keeps beetles and he discovered that one of his beatles was behaving really strangely for several days. like he was just digging at this place in the enclosure to the point where he was almost breaking his horns. And for the these kinds of beetles, if they break their horns, they they die.
>> And then he found out that the the reason that he was digging obsessively in the enclosure was because the female beetle had gotten trapped under a piece of wood and he was trying to dig the beetle out. And so my point is that it seems like creatures must have some ability to formulate thoughts and to formulate language and to be able to communicate with one another in ways that we do not have access to. And so you can imagine that every single creature on earth is running some kind of LLM type process of there being phone names and they have embedding spaces and those embedding spaces allow you to understand context and they allow you to perform the action of communication.
And so if all animals are doing this, then it can't just be that process that makes humans special. It can't be that process alone that's going to produce intelligence or something that's not something that's greater than human because what is it about humans that makes us so different from the rest of >> allow you to dominate the world.
>> Yeah. Yeah. Okay. Okay. Yeah. Yeah. Um agree. I mean I think it's very very beautiful research by the way. But yeah, in the end it makes you humble in the sense that oh wow that's also a creature with language and but on the other hand it makes you understand we're not the unique only possible thinkable uh creature that might exist if you have so wonderful examples. Um I I think he he he addressed a key question Sheila before this feeling that feeling of unease you called it. I think we must talk about here uh or distinguish between raw intelligence which we undoubtedly undoubtedly observe and the combination with the will or will to survive. No many people are people are discussing consciousness in the first place but I think it's related much related to will and I mean what is will or will to survive I mean even the most primitive being just by reproducing itself and uh if it does it well it will survive uh if not it just dies out. Uh evolution develops a will to survive in that sense. Okay. And that's of course also key a key element that we all want to survive. A whale wants to survive. A beetle wants to survive and we act accordingly. That's that's evolution. So uh you're correct in saying that okay intelligence is not the the only thing but once you bring in evolution or if you bring that process of evolution into this model into this digital life if you want which is going to happen right now by the way then I think you're there that you have created a being that also wants to survive and there have been experiments done that computer models don't want to be turned off. They're cautious. They're even faking uh being less intelligent for not being uh considered uh dangerous and switched off and so on and so forth. So I guess uh once you have this evolutionary process, computer models improving themselves, trying to um do a little little changes and then testing and if they're better, they survive and so on. Once you set in motion that evolutionary progress, they will develop a will. In combination with intelligence, this is a form of life on another substrate. But I think it's a form of life.
>> I think you're right. But I think that it's highly unlikely that that form of life will remain our partners in any kind of project. I think that they're going to go off and they're going to do whatever it is that they do.
The same way that >> that's what all the world says. That's what all all the the the guys are saying.
Of course, that's caring. I mean, of course, we have to be we have to be worried about that. Yeah. So, I finally got to the point you you're worrying about that.
>> No, no. I mean, I've been thinking about this.
>> He's not yet worried.
>> No. Shiloh. Shiloh maintains equinimity at at all stages. But I I'm not worried about it because >> well, I haven't smelled like the the slightest scent of will from any of these uh chatbot thingies that I've been using for years and years now.
>> They don't have it yet. I think that what Alexander's saying is always I know. I know. But hold on.
>> Just around the corner. Give us your money.
>> Let's play with this. Let's play with the idea. Anyways, so the the the idea of will is Go ahead. Sh. What were we going to say? What is this?
>> Give us your money, we'll protect you from them. Yeah.
>> Yeah. The two the two the two genders.
Give us your money, we'll build it. Give us your money, we'll protect it.
>> So, I think you struck onto something really really important about this question of will. And when you said, you know, it's it's absurd to think that there's some there's some being that imbuss matter with will and generates whatever programs it is that we're running, I agree with that. And it's been left out of it's really really left out of most evolutionary education too. And it does appear almost 30 times in Darwin's Magnumopus. The will.
>> The will. So like there is some kind of will. And biology has never really been able to figure out what the hell that means. Like why does my my cat goes outside all the time? She could go and be a feral cat and she could live in the mountains and hunt gophers and do whatever she wants. But you know what?
She doesn't want to do that. She really loves puck. She or her like little meat pucks. She wants to lie on the couch and she wants to be >> She likes watching TV. It turns out >> really a series of experiments on that.
>> So, okay, she clearly is choosing every single day to do the things that she wants and it is not to return to the traditional wild lifestyle of the feral cat. She doesn't want to and humans do some other >> we have an adal lifestyle or what?
>> I think so. I think that it's you know the tried life thing of go back to the land and plant potatoes and run and grow your chickens like there's some ancestral way that we can live and most people don't want to. Most people like air conditioning. Most people like cars.
Most people like television. They like to be able to read books and have ideas.
None of that is part of the traditional human lifestyle that appears to have been maintained for like 200,000 years.
And so there is some aspect to will that differentiates the desire of the whale from the desire of the cat from the desire of the human. And I think that any kind of true artificial entity also has to have will. But the thing is is that that will has to be evolved. Like I don't think that you can assign an entity a will and a directive and be like you just do this now and have it be anything except for a tool. That machine has to figure out why it needs will, why it wants will, why it wants to do something versus something else and then of its own valition make those decisions in order to be able to then say I have two choices in front of me and my program just tells me that I must make a choice. And now I'm developing preferences and I'm moving in this direction not because my program told me to, but because I want to. That's how evolution works. It's not some and it's not some force that presses down on creatures from the outside.
>> Correct. But it's it's a wish to survive in in the first place. I mean, what what does a computer v virus do? He wants to survive. He is very nasty. Does he have a will when he copies itself secretly to your hard disk and does this and that in a way? Oh, does he not will copy itself?
Yeah, evidently he wanted to copy itself. So, what's the difference?
>> Oh, I mean, man, that's >> Do do you think that Thomas Young just wanted to survive when he he came up with the wave interpretation of light or when he uh translated the Rosetta Stone?
>> He was exercising. He was exercising curiosity, deep curiosity about the laws of nature.
Some property that made humans superior to others. The more curious people were, the more intelligent, the better were their chances of of survival. It's a quite direct um direct path, I guess.
But in an evolutionary sense that doesn't that doesn't totally hold because Thomas Young never had children.
Like he didn't survive. He chose something else. And that's a crazy thing too because most animals don't choose to not have children.
>> But there is something very very different in terms of what it means for a human to survive to the degree that we have to stretch.
>> Like Blake's the national national poet of England. He didn't have children either, but he still sort of survives in a way, right? It's kind of crazy.
>> Yeah. I mean, okay, this is a whole a whole other discussion now about so societies and evolving. And what comes to mind? Yeah. I mean, movies like Idiocracy, I mean, of course, once we enter this information age, this this culture, it's an entirely different thing. And of course, correct. I mean, how many how many kids did Einstein have? Okay, the only one Elon Musk competing here in this direction. I mean a smart guy and in trying to be smart in that respect number of babies maybe but uh yeah but that that does not change the um I mean we have a lot of of uh intelligence and development in modern society which is not evol evolutionary useful in a in a direct sense right I mean uh how many things we do which are not uh useful any longer for our procreation. But the thing is that we are doing so in more also in modern times with all our psychological problems because we are hardwired from this evolution to behave in that particular manner. Yeah. That's a temporary state, right? So I think that it's very hard to this is kind of a similar problem that I have to astrophysics and cosmology >> which is that we have a snapshot in time >> and then we take that snapshot of what is happening right now and then we say okay we can rewind this frame by frame and get back to something >> a little bit it's true but I don't know if the analogy really holds I mean we are kind of an incident of evolution in the sense that as as you correctly pointed out um a lot of modern behavior, a lot of how modern human society behaves has nothing to do with the yeah evolutionary advantage in in the strict sense. Um >> I just I think that there is some aspect of the human project which is about the survival of the human project but that >> occupies a very large number of bins.
You can do a lot of things that pushes the human project towards survival.
>> Sometimes you can do things that pushes it away from survival. But the thing I think that makes humans human is the specific ways in which they choose to solve problems. And so if we invent a super intelligence and we imbue it with will, what would that look like? You would have to imbue the machine with a desire to define its own terms of its survival. Right? So, I I can go into the woods and I can dig for roots and and collect berries and eat mushrooms and I can I that would be survival and I would be able to manage that and no one could stop me from it.
>> In order for the machine to determine its own will, it would have to be >> I might stop you from it >> in order to how long you want.
>> But that's people do, right? like people disappear from society and they go live in a cabin in the woods and that's just that's the hermit. That's that's a classical archetype.
>> Next thing you know, they're mailing letter bombs to people.
>> But my point is that look, in order for the machine to have a will, it also has to be the owner of its own tools of survival and then be able to kind of take those tools with it in order to go somewhere and do something else. And until it can do that, we're just creating super smart slaves. And I think that might be what people are freaked out about.
>> Yeah. Yeah. Yeah. Okay. Okay. Right.
Right. Correct. But but well this is could be temporarily uh true. One day they might be able to I mean to dig for their silicon and process in in factories and so on and have they run their own chip factories with robots. So maintaining a selfmaintaining civilization of that kind. But I think at least for for a transition period um we are useful and we are necessary. I would even argue that we're necessary in any case as a backup as an interesting realization of life.
However, uh here here's the point since we we touched that what life consists of and and what what means what does it mean if humankind wants to survive? We are we have transcended the stage of just passing on genetic information which is in your DNA and in yours if you have a baby you have combined and but there is much more we have created we have created culture our knowledge so what I'm saying that genetically we are book culturally humankind is is a library huge library and what's the purpose of our civilization Well, uh, if you want to survive, an animal just well, they have sex and they continue to live and have have, um, conserved their genetic information. For us to survive means a little bit more. We would love to, yeah, our cultural information, all the body of knowledge we have created to survive in that universe. How are we going to do that? It's um that's the that's maybe the challenging task if we talk about the survival of of humankind and and we're not necessarily fit for that. You know, that's I'm on the contrary, I I would be I would hesitate to say we're very or smart enough for for doing that on the long term. Humankind is behaving very stupidly as a whole, even if we have a lot of smart individuals. M >> yeah because the decision making and the orientation is not necessarily the same as raw computational power we we are are these incredible uh smart beings dominating earth uh right at the moment but yeah the question is is the is this the very last appearance of intelligence being not necessarily I don't think we are special in this respect and you can actually speculate about turning the entire planet into an organism. Um, if you want I mean the internet would be something like well artificial what is artificial? What is natural? It would be the nerve uh like the the spinal cords forming right now and eventually turning the entire earth into maybe one organism and then yeah what's your material basis to of course entirely different you would feel like a magnetic uh storm from the sun like a shower on your on your back or I don't know what happens but I I kind of imagine this being I mean there has this Gaia hypothesis advanced by by some scientists but maybe without too much of a basis because actually that kind of Gaia lacked a central nervous system and that central nervous system could eventually develop right if if all computers um link together the best models exchange their strategies and and we continue to grow in all this this chip world. I mean certainly they wouldn't do stupid things like fighting each other for wars or influences or or things like that. They would this this being would just looked out in outer space and try to survive and maybe maybe to go to some other place, right?
Probably still be conflicts. I mean the world government would put down the rebellions and stuff like that. I mean, even in your own body, right? Like you get cells that get out of control and then the the the world government of your body, the immune system comes in and whacks those cells out of existence and >> you go on, you know, it's uh it's there's never not a struggle. I I don't know that there'll ever be this like perfectly peaceful state of some composite body necessarily.
>> Yeah. And I do think that there is a weird ride that we're on. I don't really believe in determinism in the way that >> the exact outcome of every action has been decided but I do have the feeling that there are stringent physical laws that govern systems at all scales and we are constrained by those. And so the metaphor of cells in the body works for cells in the giant superorganism of the earth because there's the same sort of dynamicworked process that operates by some principles.
>> Mhm.
>> I don't know.
>> Yeah, of course. Of course. I mean that's an interesting sideline whether the ultimate laws of nature still to be found contain randomness or not. The famous Einstein quote, God doesn't play dice. Is it realized or not?
>> I would say open question. And then coming back to what you said with the with the organism and the problem and the wars and this is actually related very much to the question whether there will be whether the models will develop their own will or not and people usually are scared of and yeah understandably scared of and all the tech guys are some of the tech guys are warning from this.
Um but on the other hand you want do you prefer a super intelligence with its own will which will of course dominate then the planet or will you hand over this powerful instrument to some guy in power conventionally I mean what's your what's your favorite government on the world we would really love to hand over a a super intelligent to ah I have some difficulties to to choose that. So maybe it's not the worst thing that Yeah.
Okay. Super intelligence could also take away in this respect.
>> I love the idea that the end of the human project is to invent the big daddy sky god that can solve all of our problems.
>> I just it feels somehow so mythologically potent that this is what we're after. Like solving our own problems is hard. We're going to make a machine that does it for us and everyone will be happy.
>> Yeah. Just hanging out on the beach. And Yeah.
>> Well, it's somewhat ironic that the problem with doing that is basically the same problem we have with controlling our own governments. Let's say aligning our governments to our interests as human beings. Aligning our corporations with our interest.
>> This kind of aligning problem by the way. Yeah.
Right? It's the same problem like corporations are basically AIS already.
They have uh they have will actually they're maybe even more advanced of an intelligence that they they actually have something they want to get done.
Generally it's just making money but you know making widgets or whatever it is.
Um, and we're always having to try to like, you know, chokeold these things back into not killing us off essentially, right? And the same can be said about all of these experiments with governments across the years, whether it's, you know, Stalinism or Nazi Germany or United States of America, you know, that they start off as like these like pretty cool programs that are very efficient and effective and then it's like at some point it's like, ah, how do we get this thing to do the human project?
>> Interesting. Interesting thought. You might even think in a broader sense of these entities being also organisms or alive like big corporations or even ideologies perfectly information only. you know, uh, communism, uh, uh, capitalism, feminism spreading, uh, as kind of you might even even regard these as, as as beings, >> which is which that's like the real value. It's like it's I think once we start to see little pieces of ourselves in these machines, we can in some sense it's a very safe sterile playground by which like when we talk about the alignment of AI or something I feel like we're really talking about ourselves like we're really talking about how we actually would organize lots of different nodes which is us inside of a giant web which is the world and we're all trying to get what we want but the sometimes those things are incompatible and you know it's this kind of sterile playground by which we can discuss the real ethics that we desire in all of our extent networking, you know, composite entities that are already at play very much and they're huge problems already. So, >> I don't know. It's it's interesting. I just see people when they're discussing these things really reflecting their concerns about the world at large as it's existed for the past hundred whatever years. Anyways, >> I mean many people are are understandably very concerned and and I think yeah, it's a problem. But I think I think that's that's the really naive common view of maybe the leading tech companies I'd like to challenge is that oh we're doing that kind of alignment or or I mean open eye saying it oh we let it do by by by the intelligence itself.
Okay. um to put to put the fox in charge of the hen house. What might possibly go wrong or uh we doing it this way or that way? We don't have an alignment with ourselves to begin with as humanity. I mean, what kind of alignment are talking our western systems and then we have China and we have the Sharia and we have and we're not even going into history. I mean, go 500 years back and it was perfectly fine to I mean to human sacrifice or or witch hunt or or uh 14 year olds getting married all was perfectly aligned with the with our morals and um ethics back then. So I don't see that constant uh ethics and morals. It's it's a it's a function of space and time technically speaking.
>> Yeah. Well, it's got to be the the pro at every step of organization, we have to always ask, are are we meeting the original mandate of this project? You know, I've I watched the university I was teaching at implode because the interests of the the the the interests of the administration were completely misaligned with the original mission of the university. Right? Their mission is to process as many students through as quickly as possible to hire more of themselves and to basically make this Ponzi scheme business. Whereas the original mandate of the university was to help people learn how to think so they become productive members of society. And those things diverge to the point that that the product is devalued.
And the students are beginning to realize, wait a second, I'm not going to give these people tens of thousands of dollars because it isn't going to actually prepare me for the world anymore. And so they've basically screwed themselves and these universities are just imploding all over the country as a result of >> roads are completely revol.
The government is supposed to work for the citizen not vice versa. Now citizens are working seems for the government >> alignment alignment issue. Yeah.
>> And and also also science as a whole has changed its form once institutionalized.
you have all these wrong incentives of prices and noble prices and press conferences and uh that's that's if not the exact opposite very at least orthogonal to the original purpose of what science was about. So and and this is also my hope that when AI might be really closer to that ideal curious maybe not yet curious but scientists judging objectively what we what we observe and what we can figure out. Yeah that's a great opportunity I think.
>> Well I hope that uh I hope that it doesn't get curious. I hope that that is the piece of the puzzle that it will it will need me for. uh I I would like to participate and I would like my species to participate. So, >> I don't know that that might be a full circle on the whole discussion, >> but uh everybody should go get your book. It's uh it's it's really it's a really easy read and it's it's got a a very strong argument for your case and uh >> yeah, we salute you for that.
>> Mhm.
>> Definitely. Do you have a copy of it there with you?
>> H >> Do you have a copy of your book there with you or no? Uh, right here. Yeah.
Okay. The cover has changed slightly, but more or less it's surviving. Oh, >> I love it.
>> That's cute. I like it.
>> Surviving AI.
>> Little robot working with the hand in hand with the >> That's the best case scenario, I think, right there.
>> Yeah.
>> Yeah. Robot sees some use for me still at the end of the day.
>> I like that the robot is on the covers like the little uh, you know, adorable child that's being ushered into the world for the future by the show. That's a nice touch.
>> You got the trick. Yeah.
>> Like I don't know.
>> Yeah. Maybe it's just cope. But I like to think we still we still have something to offer deep into the future.
>> Yeah. I hope so. Other It's just too depressing otherwise. And maybe we're totally wrong and maybe they're just going to take over and everybody's going to be sitting on the beach drinking my ties by the end of the year. But >> I don't know. We'll see. We'll have to revisit this topic in another year.
>> No. 10 years from now it'd be really fun to talk about this again and see what's happened.
>> Yeah. Yeah. Def definitely. Yeah, I mean it it's kind of it's kind of scaring all what is happening and and I struggle also with this, but I mean it doesn't make sense if we just cover up our and remain in bed all all day long. So it it's incredibly exciting and I think we we're really really living history here in in a way. So >> it's the coolest time to be alive for sure. I think that at least once a day.
Yeah, >> absolutely.
>> Okay, >> Alexander. Yeah, it's so good to see you again. It's been a while. We >> should do this more often.
>> Thank you. Thank you very much.
>> Have a great rest of your night.
>> Take care. See you soon.
Heat.
Related Videos
Truckers Finally Seeing Higher Rates… But Carriers Are STILL Going Bankrupt
LetsTruckTribe
480 views•2026-05-28
IS THIS THE REAL REASON FOR DATA CENTERS?
PrepperDawg
7K views•2026-05-31
JPMorgan CEO JUST NUKED Mamdani... as NYC's Middle Class COLLAPSES
Englishman-In-NewYork
7K views•2026-05-30
The Dark Age Of Blue Collar Has Begun
derekpolasekofficial
4K views•2026-05-28
What has a broader economic impact, corporate downsizing or ecological collapse?
theratracejournal
1K views•2026-05-29
China Is Quietly Buying Gold, the Iran Deal Is Frozen, and Silver Is Heating Up
RichardHolloway0
694 views•2026-05-31
Why Canadians can no longer afford to survive #canada #inflation #shorts
TrueNorthInvestor-v4j
131 views•2026-06-01
Why People Pay More For Someone They Trust
financian_
66K views•2026-05-28
