Where Does Genius Come From?

Michael Levin 00:00

In the last couple of years, we've made novel living beings that have never been here before. So xenobots and anthrobots. These are these are novel creatures that have never been on Earth before. They're made of living cells, but frog and human in one case and the other, adult human cells.

Michael Levin 01:06

Yeah, thanks for having me. It's good to see you.

Michael Levin 01:27

Sounds great. Yeah. Yeah. Tell me, tell me about it.

Michael Levin 01:59

Well, I'm not going to claim that I have one definition that captures everything that we want to know about intelligence, but for the purposes of our experimental approach, I dr I define it as competency in reaching it basically in problem solving. I focus on the problem solving aspect and degrees of ingenuity in reaching one's goals as circumstances change.

So we look at a wide variety of systems, not even just when you know you say species, but the but we look far beyond the living things and we look at cells and molecular networks and slime molds and inorganic materials and populations of I mean all sorts of different things. And we look at different ways in which they solve problems and they achieve their goals in navigating some kind of problem space. So I don't think this is everything. This doesn't capture play and exploration and the some other things you want to but for our purposes it's plenty.

Michael Levin 03:04

Well, I'm open to the idea that there are other forms of intelligence that have nothing to do with go pursuit of goals. So for example, exploration, creativity, play, those kinds of things, right? That's great. I'm sure those are very important. Those are harder to quantify. And so for us, because you know we already have a fundamental problem, which is that people are very resistant to study and to find intelligence in unfamiliar substrates.

So if I start getting into things that we can't quantify in an objective way, then it becomes very difficult to convince anybody that we're doing more than some sort of, you know, wishful thinking or you know, some sort of animist approach where you just sort of paint a spirit on every rock and so on. So we try to just for the experimental work, we try to stick to things that are easier, still not easy, but easier to quantify, which is goal directedness. But no, I don't think that captures everything of interest, of course.

Michael Levin 04:11

Well, what do I actually do? I try to do research on things that interest me. So deep questions that about life and mind and meaning and things like that. I try to figure out ways to ask nature about those things and to do experiments and then try to interpret the answers in some way that allows all of us to have a better life. I think that's what we do.

Michael Levin 04:53

The first thing I would say is that all of these things are empirical questions. In other words, I don't think you can answer any of the sitting in an armchair. So people often say, Well, that's just the whatever. It can't. And I say, Have you done the experiment? Do no, it can't. And so the first thing I would insist on is that whatever we say about this is the result of experiment that you tried and to the whatever extent that and of course intelligence is I think a two-way IQ test.

So it's very easy to give something a low score of intelligence because we're not smart enough, we're not creative enough to know what it's doing. So we have to be very humble about those things. The only two things I've I insist on strongly is that one is that it's an empirical fact that we try to discover, and two, that it's a spectrum. I don't think it's a yes, it is intelligent versus it isn't. I think the question is what kind and how much. And I agree with you that I don't think any one particular metric is going to be sufficient. I think probably it will be a you know, some sort of a complex sum.

Descriptor. But we can ask the question, does it go all the way down? In other words, in this in the spectrum, what's a zero? You know, what earns a zero on that spectrum? And if we want to think about that, I would say there are two criteria for being on the spectrum in the first place. The first criterion I think is that you have to have some degree of goal directedness. So simple homeostatic kinds of properties where we can identify some kind of state that the system is working to achieve.

That's A. And B, another thing that's good to be on that, on that spectrum is that the information you need in order to meaningfully relate to that system, for example, to know what it's going to do next or to communicate with it or whatever, that information goes beyond adding up the immediate pushes and pulls on that system right now.

So if you know exactly what's going to happen just by figure by adding up everything that's happening to it, you're looking at a billiard vol of some sort. And most of the tools of behavioral and cognitive science are not going to be of use to you. And we quantify that with this notion of a cognitive light cone. So the cognitive light cone I define as the size of the biggest goal state that you can pursue. So we're talking about the amount of space, but also amount of time going backwards and forwards that's relevant to the actions that you're taking.

And if it's and if that horizon exists is extremely short, you were probably not dealing with us with a significant intelligence. And then it's just an experimental question of all the tools that we have for dealing with intelligent systems, how many of those tools are applicable? So if we take that definition, if we say those are two simple things, the minimal things that you need to be on the spectrum, then we can say, is there anything in our universe that doesn't meet those criteria? And I actually think that.

You know, if you even look at simple particles, you know, subatomic particles, I think least action laws give you something akin to goal directedness, so error minimization, goal directedness, that kind of a thing. And certain kinds of and I'm no physicist by any means, but my understanding of what we've learned about the quantum world in the last however many decades is that you also have situations that where traditional billiard ball causation fails, where you simply cannot figure out what the system is going to do by adding up all the local pushes and pulls. You know, randomness is not a high level of intelligence. We don't want to be random. We want to be intentional. But nevertheless, I don't think it's zero. I asked a Chris Fields once who's a who is a physicist in my center and I said it, does is there can you imagine a universe in which there was no least action laws? And he said, well, only if nothing ever happens. So presumably there are static worlds that are at zero, but I don't think that's our world.

So I would say that in our universe, it probably does go all the way down to some extent, but the degree to which the tools of behavioral science are helpful to you are, you know, rise sharply as you enter the biological world and so on, and they're sort of moderate use, not zero, but moderate use below that.

Michael Levin 09:05

I hope you haven't seen me drawing any lines because I don't I hate drawing lines. I think drawing lines we draw lines and then we as you know, where things are convenient for us and then we sort of start to believe that these are real lines. Now the thing is, I don't say much about consciousness yet. I will, but I haven't said much about consciousness yet at all. And I and partly it's because I think it's a it's an entire it's related but an entirely different kind of question than intelligence. Because intelligence

Michael Levin 09:35

You agree?

Yeah. Intelligence and other behavioral competencies are good third person observables that we can all agree on the experiment, we can all agree on the outcome, and they should be, you know, it's standard kinds of science. So questions of consciousness, I think, are different. And we can talk about what I think it means, but I think many people who say they're working on consciousness are really working on behavior or physiology. They're not actually working on consciousness. I think it's very difficult to work on consciousness. I suspect.

That degree of consciousness tracks intelligence to some extent, but I don't think they're the same thing. I don't think that they're necessarily tied, you know, sort of in a one-to-one kind of relationship. And I'm overall very suspicious of what people call unconscious processes, whether in the body or outside of the body, for the reason that they're all calibrated against human verbal reports. And that's an N of one.

And when somebody says, okay, you know, whether it's subliminal reading or whatever it is, well, that was an unconscious process, I say, How do you know? Well, I asked the subject and he said he has no re no recollection of it. Great, you talk to the part of the being that has access to language, and that part didn't wasn't conscious of it. Cool. But how about all the other stuff? And I th that's within our bodies and our brains. And I suspect I think the same goes for example, anesthesia. So I was having a discussion with somebody who said, Well, I anesthetize the patient. And my feeling is, well, you anesthetize the one with the language on the wallet, who may or may not sue you afterwards. So that's great. So he was anesthetized. But how about all of the other potential minds in that system?

That don't have access to the language and that you we don't have this is why one of the reasons we're doing in our lab is building tools to communicate with cells and tissues and molecular networks, all kinds of weird stuff, because I think we have a massive degree of mind blindness. And mostly because we're fixated on language and we are fixated on this idea that there's one individual inside a human body. And as long as we talk to the sort of left hemisphere, then we're good. We're getting the direct story. And I'm very suspicious of that.

Michael Levin 12:53

Yeah, I mean, just you know, in the context of the things that we study. So we study problem solving, memory, learning, decision making, goal-directed navigation, starting from molecular networks, cells, tissues, organs. So as far as I'm concerned, the a biological body is already an ecosystem of an immense number of different intelligences who all project their competencies in different spaces. So for example, we're all used to brains.

Solving problems in three-dimensional space. So using muscles to move animals around in 3D space. And now as humans we like linguistic space and maybe chess space and things like that. But those are you know those are the kinds of spaces that we've learned to recognize intelligence in. But look, cells and molecular networks and things have been navigating other spaces long before any of that stuff evolved. And in fact I would make the argument based on evolutionary homology that all of the tricks that brains use are basically speed ups and expansions of the cognitive light cone and then projection into new spaces of tricks that evolution and probably even before evolution, other aspects have been using long before brains and muscles appeared. So we see cells navigating c 20,000-dimensional transcriptional spaces to solve new problems that they've never seen before. We see metabolic spaces, physiological spaces.

But the scale of these things makes it hard for humans to recognize them. They can be very small or very slow or very fast or you know, look large and so on. Then it's very hard for us. I think that, for example, if we had evolved with a primary internal sense of our blood chemistry, the way we have taste and smell and so on, if we could feel the different parameters of our blood chemistry, I think we would have no problem recognizing that we host a bunch of symbionts like our liver and our kidney, and that these creatures navigate these spaces to keep us alive every day and we would be able to.

Actually visualize them doing it in these spaces, but we can't because we're obsessed with vision in the sort of three 3D world. We have no idea what our organs are doing. So having said all that, given that I think we are this kind of ecosystem of numerous intelligences which are which are nested, they are probably overlapping in the in terms of the physical hardware that they supervene on. They're interpreting each other's activity all the time. They're hacking each other in various ways. They compete, they cooperate.

Th all of these things have agendas and different degrees of competencies and different spaces. And so, you know, when you get a verbal report from a human, that's like part of the story, but it's not remotely the whole story.

Michael Levin 15:47

Yeah, in many ways. We've seen things. We right. We have assays that for first of all look like an IQ test because it's exactly those kinds of things. You know, you're given some affordances and then a problem and it's a you use this to solve a problem you've never seen before, right? Like the thing where they give you the, you know, the thumb tax and the candle and the, you know, and to solve this like we see things like that.

And then just navigation of spaces with barriers where this is you here you are, there's your goal, here's a barrier you can't get through. What are you gonna do? And all kinds of interesting things that human subjects couldn't even begin to solve.

Michael Levin 16:53

Yeah, so let's work it backwards from the psychology end, I guess. One of the cool things is that there is remarkable symmetry and invariance between neuroscience and developmental biology. We pretend there isn't by having different departments and different journals and different funding sections and different classes and so on, but actually the tools can't tell the difference.

So we take all we, you know, we appropriate all the tools of neuroscience, both the practical stuff like optogenetics and you know, ion channel mutants and neurotransmitter drugs and so on, and all the conceptual stuff like you know, perceptual control theory and active inference and all of these kinds of things. And we use them on things that aren't brains, yeah, other cells, other cell types, other situations in the body, regeneration and development and so on. And the tools work perfectly well. They can't tell the difference. It's all the same stuff because every cell in your body has ion channels.

Most cells in your body have electrical synapses, tons of neurotransmitter machinery from the time you're a single cell embryo. So these are not I don't like I actually don't think neuroscience is about neurons at all. I think neuroscience is about cognitive gloom. So, what I mean by this is this question, the like the most beautiful thing about neuroscience is that it addresses the following question: why do you know things that your individual neurons don't know? How do you have goals that your individual neurons don't have? What's the scale up here, right? And we've been told that the mechanisms that implement that scale-up, the cognitive glue in the brain, is electrophysiology, that basically your neurons form electrical networks, and these electrical networks can operate at a higher level. They know things that the individual cells don't know. They project goals, they have you know, memories, plans, representations, hopes and dreams, and so on. And so then we can ask the question, okay, amazing. When did this capability evolve? Well, it turns out that the earliest versions of something like this were around the time of bacterial biofilms.

The first microbes were using electrical networks to coordinate information across space and time to solve metabolic problems, eventually solving using them to solve anatomical problems. What's an anatomical problem? Well, how many eyes should I have? I need to remember. How many eyes do I have now? Zero. I need to have two. So how do I get from here to there? And what happens in the meantime if somebody removes one of them or if there's an extra one popping up? What do I do? These are all anatomical problems. So it turns out that.

Long before your neurons were there to think about getting you around in 3D space, the exact same machinery was used by all your body cells to get you around in anatomical space. You start off as a single cell, you have to become a giraffe or a human or a snake or whatever you're gonna be. And this is not just, contrary to what we're taught, this is not just some sort of feed forward unrolling of the genome. It is an active navigational process with lots of cool context sensitive problem solving capacities, learning and all that kind of stuff.

So if we ask the question, what is the cognitive glue there? It turns out it's exactly the same. It's the ability of groups of cells to form an electrical network that can store larger scale goals than single cells can alone. So the cognitive light cone expands and projects into a new space, whereas single cells can store metabolic and physiological goals, pH level, hunger level, things like that. Groups of cells make a large network that then can store grandiose goals like here's the shape of my limb and what happens if that if you're if I'm a newton that limb is amputated, how do I get back to where I need to be? And when do I stop? Right. When do I reach to the my goals and then I can stop. So those kinds of things. So yeah, so that's bioelectricity. Bioelectricity is just understanding the broader role of electrical networks in scaling cognition and scaling the scope of one's goals and one's competency to pursue those goals.

And to the extent that we understand that, we can do a couple of interesting things. And we've done this in our lab over the last 26 years, I guess. One is we created the first molecular tools to read and write that electrical information out of cells, like neuroscientists try to do in the brain. And so the first thing you can do is read it and thus get a preview of what is the system thinking about building in the future. We have the ability to read the electrical information and thus get a preview of what the system so looks so kind of like mind reading, basically like neural decoding.

Try to understand what the system is planning on building. So the electric face shows us a pattern of what the embryonic face is going to look like long before any of the structures are forming or the even the genes are turned on. You can look for birth defects, the all that kind of stuff. Then, of course, even better, you can incept novel memories into that what you using the to the ion channel drugs and optogenetics and things like that. You can write new information into the tissue.

So you can say.

Michael Levin 21:37

We haven't we haven't done human yet. Well, we've done human in vitro. We've human cells in vitro, yes, but the but all the in vivo stuff is animal models. But ultimately the idea is to repair human birth defects, right? To regenerate human organs, aging. I'll tell you a cancer story in a minute. So the idea is that you can write new information. And if you have a flatworm that normally says, I remember I need to have exactly one head, you can write in information that says, Nope, you should have two heads.

And or we can take a tadpole and say, you know, you should you really should have an eye on your butt. And they will, if you write the correct information, that is their memory of what they're supposed to look like. Don't touch the genome during any of these things. The genetics are not involved. The genetics at the hardware, and as we all know, certain hardware, if it's good enough, can host all kinds of interesting software. And it's this stuff is highly reprogrammable. And I'm not claiming that the standard formalisms of software and hardware are sufficient to tame the biology because they're not.

But there's a few parts of them parts of these things that map nicely and reprogrammability is one of them. So yeah, so that's what we can do.

Michael Levin 24:14

Yeah, super, super interesting and important. Okay. The first and foremost thing I want to say because I'm next going to talk about a lot of ways in which our current formalisms I think are deficient and are missing something very important. And I'll make a guess as to what that is. But the very first thing I want to say is that what I'm not making is an argument for some sort of mysterianism. When I talk about these things being deficient, what I'm not saying is look, there's some magic out there and we're never gonna figure it out. And so let's just assume right.

Michael Levin 24:44

There you go. That's it. That's exactly it. The what the process of science, not the current content of whatever the science of the day is, but the process of rational inquiry. And so okay, and so I want to be clear on that because the very weird, the very weird framework that I'm about to tell you about is not meant to be it in fact, I think it's the opposite. I think it's meant to move past some impasses, as you just pointed out, and there's some other things. I think emergence, this notion of emergence is really putting the brakes on things.

Michael Levin 25:16

Yeah. I don't think it just even begins to scratch the surface of what's actually going on. You know, it's just early step. So the way I think about these things is the following. We have, both in biology and engineering, we have some degree of expectation of how much work it takes to produce something. Let's say something of competence, whether it's a robot, an algorithm, a machine, a an organism that's good at doing something rather than has specific features where it's do doing something cool. We kind of have three ways that happens. And everybody loves this idea of no free lunches. No free lunches means that you have to put in the effort. You don't get anything for free. And in fact, the one thing that Darwin and Paley agreed on, they didn't agree on where the source of the order was, but they both agreed that order is extremely rare, the needle in a haystack thing, and that you need an explanation for it. They were in total agreement on that.

That if you find a watch or an animal outside, you have to come up with some sort of very difficult process by which that came to be. In modern science, we have basically three ways of paying for that kind of competence. One is rational engineering design. So if you already know how to do something, you write an algorithm or you build a machine, but somebody has to already know how to do it. That's where it could, that's where this thing came from because somebody knew how to do it and he wrote the algorithm. That's one. And that's why engineers are you know highly paid and so on. They know how to do it.

Number two is okay, evolution, natural selection. Nobody knew how to do it, but that's okay, because we paid the cost by generating trillions of variants. We kill off everything that doesn't work, whatever's left, that's going to be competent. Okay, so that's great. Nobody needed to know how to do it in advance, but you still had to pay the cost by generating and testing a massive number of variants. Not to mention you had to have a self-replicating system to start with. That's a whole other thing.

The third thing is learning. So you have a system, it doesn't know anything, but it had plenty of contact with the problem before, took its lumps, and eventually now it got good at what you're doing. Okay, no evolution, no, no, no design, but you paid the cost by having this thing confront the problem and try again, right? So those are the three things. And whatever we see something complex or competent, we think we can tell a story based on those three things of where this came from.

So what I'd like to do is to define the size of the discount of that lunch. Let's say it's not free, but the size of the massive discount that you may or may not get is the delta between the effort you put in and what you got out. And I think the and I'll show you some examples where so

Michael Levin 27:53

Well, what makes it free is if you put in very little effort and you got it and you got a massive outcome. Right. Yeah. So in your field the kind of thing that you were just telling me, it's like when it's the mismatch between expectations and what you got. The kinds of things that yeah, somebody got whacked on the head and now they're a mathematical product. Like it's not common, you wouldn't suggest that as a method, but it does happen. There are clinical cases. Like

Michael Levin 28:27

Yeah, I'm gonna get to that as terms of the navigation of this platonic space. So yes.

Michael Levin 28:39

And go on. Super. Yeah. And I'm interested in any examples that you have, because I've been looking these up, but obviously it's not my expertise. I mean, the other example where you get stuff like this in neuroscience are the people with massive amounts of brain deficits, right? Where the rare hydrocephalus patients are something where like, okay, they have less than a third of the cortex of a chimp, but they have normal human or above normal intelligence, right? So the thing with those kinds of things is that even

Michael Levin 29:22

Yep, exactly. Or that case from what the seventies, I think, about the twins that quoted prime numbers to each other, you know, without a math education, like there are these crazy things. And it's not that you can't shoehorn these into neuroscience. Well, redundancy, so you can sort of wave your hands about these things. But the standard models in the field don't predict that this should happen. Okay. And it's surprising because that is not the expectation where we say, well, you need this brain area for this and that brain area for that.

And by the way, we still, of course, don't know what is the cognitive density of brain tissue because somebody has 63 personalities in the same brain, and you know, and somebody else has very little brain and then the normal personality, like we still don't understand what the carrying capacity of it is. So we can start to try to understand what is the delta between the effort you put in and what you actually get out. In our lab, we have two sets of experiments that we're doing on this: biological and computational. The biological experiments are in the last couple of years.

We've made novel living beings that have never been here before. So xenobots and anthropots. These are these are novel creatures that have never been on earth before. They're made of living cells, but frog and human in one case and the other, adult human cells, not embryonic human, but adult human cells. So we can take cells out of your trachea, people have biopsies and then they sell the cells and they make they we put them in a Peter dish and they make a little creature that runs around by itself and it does all sorts of cool stuff.

So we can ask the very simple question, when did we pay the computational cost to design these things? Normally you would say selection, you know, you eons, where did we get the frog? You know, what's responsible for the frog, for the properties of the frog? Selection. The genome's been bashing against the environment for millions of years, and therefore you that's how you paid for the frog. Okay. How did we pay for the xenobots and the anthrobots? There have never been any xenobots and the anthrobots. There's never been selection for it. We didn't program them.

We have no idea how to program any of the things that they do. We didn't design them. We didn't select them. We didn't train them to do any of these things. Where do these things come from? So I think this is really interesting. The question of Yeah. Yeah. Well, where do the properties of novel beings come from? Now, now on the one hand, it's interesting. Or novel abilities. Novel ability, exactly, right? So we're all going to be living with different kinds of cyborgs and hybrids and different kinds of beings soon. Where do their properties come from?

When you have a three-year-old who sits in front of a piano, everyone's better than I do, like where the hell did that come from?

Michael Levin 31:51

It's an empirical question of bait when you can't blame it on evolution, where do these things come from? So okay, and then that's the biology side. Now, the thing about the biology is that it's very hard to prove anything because you never know all the mechanisms. There's always some kind of you know, the quantum biology or something underneath that you didn't know about.

So we complement that kind of work with minimal computational models where we know all the pieces. We can see the algorithm, we know everything that's there. It's a simple deterministic machine. And yet, as it turns out, and we've had a paper on this recently, we're gonna have a bunch more stuff coming out this year. There's a ton of work coming soon that shows that even simple so-called dumb machines, which I don't really believe exist anymore, are subject to this kind of free lunch. The you get things that you did not evolve, design, it's not in the algorithm, you didn't design it, you didn't train it, you didn't evolve it, and yet it has access to it. So the way I so now we get to the genius portion. So the way I try to conceptualize all this is I think this is all part of one phenomenon. And that phenomenon is a spectrum of inspiration. If the question is how much inspiration did you get? And I think and I think it goes all the way down. I think even simple machines get some of it, but I think, you know, human geniuses get more of it and whatever. And I think it looks roughly like this. And just imagine you know, and again I'm sure you have plenty of other cases to talk about, but like what we can imagine is step-by-step problem solving. So you navigate that space of ideas laboriously, conventionally, step by step, you pay for every step that you take. I'm walking across that landscape. Then you have someone like von Neumann who looks at a problem and goes, Well, it's this and then done. He didn't need to do all the individual steps. He

Michael Levin 33:50

Sort of jumps across that landscape. Then you get the then you get somebody like Ramanujan who says, Yeah, I don't know, a Hindu goddess whispered them with these theorems in my ear at night and that's it. Right. So he, you know, he just flies across that landscape.

Michael Levin 34:05

Yeah, yeah, yeah. I think exactly. And so I think we can quantify all this stuff by the way that it navigates this kind of space and by the way that you get surprises. I call them ingressions. They're only surprises because they don't need to be surprised. They're only surprises because we don't we're our formalisms for this stuff are limited.

Michael Levin 34:32

The surprise is a my goal is to remove the surprise element. So it shouldn't be a surprise. It's a surprise now because our current formalisms don't take this into account. Once we have a proper and I'm gonna tell you what I think is going on here.

Michael Levin 34:50

An ingress is when you have built a physical interface that could be an embryo, a robot, an AI, a an animal, a biobot, whatever, any a machine, anything. You've built a physical interface of some sort, and that interface becomes the host to a variety of competencies of different kinds that are not easy to discover. You need to be paying attention. Like we found some in this thing called bubble sort, which every computer science student has been studying for 80 years. Nobody noticed this before. So you have to have the imagination to notice these things. But having noticed them, you what you find is that some percentage of them you think you can explain in conventional ways. Like yes, the yes, we wrote an algorithm for this, we trained it to do this, we evolved it to do this. But then there's a bunch of other stuff that you can't explain in any of those three ways. There's something else, there's another sort of source. And actually, once you see that source, then you start to ponder, wait a minute, some of these things that we thought we had conventional explanations for, we assumed.

Is that really where those come from too? We're not sure anymore.

Michael Levin 35:57

And I think it would be very cool to write the to write a nice review of that aspect, right? To really look at what the cases are. I mean, we wrote a

Michael Levin 36:06

Lot, like yeah, Karina Kaufman and I wrote a review of some of these missing brain cases just as a yeah, you know, as one thing. But as you study, there's like tons of these other examples. So what is actually going on here? And for that, I just have one more piece to add to this, which is this notion of the platonic space. So and when I talk about this.

I am not trying to stay close to Plato's original theory to whatever extent we even know what he really thought. I'm not trying to stick close to that. But there's a reason why I use his name for it. And then and the reason is because a lot of mathematicians immediately know what I'm talking about. And here's the deal. So physicalists, people who believe there's one realm and physics is it. The physics is in charge of that realm, make the following claim. All important facts are fundamentally facts of physics. That's that would be the claim. And so now the question is, are there facts that are not facts of physics. What does that mean? Facts that are neither discoverable by nor changeable by things you do in the physical world.

Michael Levin 37:19

Yes, now that's not the cr that's not unique to me. That I mean Pythagoras knew this and people in the East knew it long before that. That's not the new part that I bring. I br I bring something else. Then the new part we already knew, I th I think, although of course some mathematicians go a different they have a different metaphysics about this. But they already knew that we have certain facts. For example, for example, here's a fact. There are formulas for solving quadratic equations, right? We all learned it in high school, the quadratic formula.

You can do that for equations of grade three and four, but it's provable that you can't get it. I think five and above or six and above, they just don't exist. Now, okay, that fact that it's doable for all of those things, it isn't doable for equations of higher for higher degree. What would you do in physics if you're a god at the beginning of the universe and you could tweak all those all the constants around the Big Bang? What would you do to change that fact? What would you do to change the digits of E?

No there's nothing you can do, right? These facts, the fact that the prime numbers have the certain properties and certain symmetry groups do these things you can't just fire all the mathematicians and hope the physics department discovers these things. You can't tweak anything in the universe, in the physical universe to make these things different. So I would claim we already know that there's a space of truths.

That is not physics, and yet it affects what happens in the physical world in terms of biology, in terms of physics. If you want to know something about particles, you say, well, why do the fermions do this or that? Eventually you get the answer, it's because this the symmetry group you know has this and this properties. Okay. If you want to know, hey, why do the cicadas come out at 13 and 17 years? So the predators don't time their emergence cycles. Okay, but why 13 and 17? Because they're prime. Well, why are 13 and 17 prime? Go talk to the math department.

So all these questions always end up somehow in the math department, all right. Eventually it always boils down to that. So right, so these things determine in an important way what happens in the physical world, but they don't but they themselves don't come from the physical world. And so that we that has been said a million times in the past, although people forget and pretend

Michael Levin 39:38

So I'll tell you this the minimal version, which a lot of mathematicians already buy, and then I'm gonna say the crazy version, which I think is more true. Yeah. The minimal version is simply this. We have the physical world governed by things you discover in physics, and then there's a set of facts that are discoverable, they are coherent, they are structured, it's not a random grab bag of stuff. People are studying it. There is a map of mathematics that people are doing with the making where all of these things are sort of mapped out and stuff

Michael Levin 40:09

No, I think black holes are part of physics, as far as I know. I'm not a physicist, but as far as I understand, black holes are a phenomenon in our physical universe. Things like you know, things like the digits of E, right? The natural logarithm, the specific digits of E. There's nothing you can do in the physical universe to change you couldn't yeah, you know, people say, well, math is just it's a it's an invention of humans. Great. Do you think you could invent it so that the natural logarithm was seven instead of two point seven eight, whatever?

You really can't.

Michael Levin 40:47

Yeah, yeah. No, it's a good question. If I forget the br bring it up again, it's a good, it's a good point. Where we are so far is basically where Pythagoras had us a couple of thousand years ago. This is like that part's already dude. I'm not the first person to say this. Here's the weird part that I'm adding to this. If we already know that there's a space of truth, specifically patterns, you know, things like certain fractals have a specific shape, not a different shape, but that specific shape, right? And so on. And that affects the physical world.

What's the guarantee that the patterns in that space are all static, unchanging, unidirectional things like the digits of E and pi and all of those things? What if, what if there are other patterns in that space that would so we knew so let's say we know in that space of patterns there is one layer that is amenable to the tools of mathematics. These things are static, they don't change, they don't at least we think they don't change, they don't go anywhere.

You know, I could tell you that the speed of light changes over cosmological time, and you'd be Yeah, well, okay, I mean, fine. If I tell you that the digits of pi change over time, you'd be like, that's insane. That's what everybody tells me. So you have one layer that is this kind of static, like minimal, basically like the rocks of that space. They're not going anywhere. It's the you know, they're it's a unidirectional thing. Nobody's changing them, they're just but what if there's other things in that space.

That we use different names to recognize that are not math anymore? What if there are other patterns in that space that are not static patterns like the static digits of E, which are, by the way, are not, you know, that's an algorithm, really is what it is. What if some of these other patterns are things that we would recognize in a behavioral science department? For example, their patterns of habituation or sensitization or anticipation or delayed gratification or outpainting or you know the or grammars or that kind of a thing. What if we would recognize some of these patterns as kinds of minds?

Michael Levin 42:54

Absolutely possible.

Michael Levin 42:58

Maybe I mean Plato did say, you know, there's beauty and truths and that, you know, this kind of stuff. I don't know. But my point is so here's what I'm bringing. I think physics is what we call the things that are constrained by those patterns. Biology and cognitive science are the things that are enabled and facilitated by those patterns. Wow. And I think that math is really be you know, you ever seen that it's kinda like a pyramid that has you know, of physics on the bottom and then there's chemistry and biology and I think the thing's upside down. I think everything is behavioral science from the word go and it's just

Michael Levin 43:41

I thought yeah, I thought you might. I think it's all behavioral science. And I think math is the behavioral science of one layer of that space that is the low agency stuff that it's amenable, not zero. I don't think it's zero, but I think it's the low agency stuff that's amenable to formal models like proofs and things like that. But there's other patterns in that space that are things like different kinds of learning and so on that ingress into all sorts of things, like we've shown learning and memory in models of gene regulatory networks, for example.

And so there are other patterns in there that are completely recognizable to any behavioral scientist. And so what I think is happening then, if you build some kind of physical interface, whether it be a machine of some sort, an embryo, a body, whatever, you are what you're really building is a bunch of interfaces to different patterns from that space that are.

In some way, a good match. This is what we're still working out what the right vocabulary for that is. But basically you're making interfaces through which these patterns come forth. And the reason they're surprising is because we have this idea that we make these things. We build them as engineers, maybe we evolve them and search for them, maybe we train them. I don't think we do any of those things. I don't think intelligence, I don't think minds are made either biologically or by AI. I don't think you make them.

I think what you make are interfaces and when you make

Michael Levin 45:11

Yeah, that's it. That's exactly right. You make and what you make are opportunities for some of these patterns to aggress into the physical world and to you know interact with other with others.

Michael Levin 45:26

And a

Michael Levin 45:27

So it's fundamentally so the funky thing is that it's fundamentally crazy and new, right? For science, like this is wild. But also it's fun I mean, I also get emails from people that say, Well, you know, the ancient sage or whatever said this in three thousand BC. Like yes.

Michael Levin 45:46

I think that is exactly right. So inspiration, genius, like all of this stuff. I think that's what it is. And I think the real question is what degree of, you know, what was your interface? And what degree of inspiration does your interface allow? If you make a biological, you are accessing some really interesting, rich patterns from that space. If you make a human, if you make a human that's done a little bit of practice, maybe even better, but it's not the practice doesn't create it.

And this is why you have these crazy and so what I've been collecting are examples, descriptions. And of course these descriptions don't necessarily mean that's how it is, but I'm still interested in how people perceive these things of creatives when they download a symphony or a right. So there's a poetess, I forget her last name, Ruth, something or other, where she said that you know, she said, Well, I was I work out in the fields.

And every once in a while I hear this incredible noise, like a locomotive running towards me. And she said, and I start have to run to the house and get a pa a p a pencil and paper. And if I'm late, it runs past me and goes to another poet. And if I manage to grab it, I guess she says I grab it by the tail and then she writes it down backwards, back to front, she because she's like pulling it in, then she's able to like she's able to catch it. Right. So there's tons of this stuff, where people say, Yeah, I just sort of like, you know, I tuned into it.

Michael Levin 47:22

I think you should definitely write an overview of it, like examples, like good examples of this kind of stuff. And I think it would map very nicely to some of the stuff that we're doing. Right. Michael Jackson, I happen I saw a quote from him where he said, I have to take these things very seriously because if I don't write them down, they'll go to print. You know, I think all of these are fundamentally the same phenomenon in different guises. It's this issue of tuning in to patterns that are there. Some of these are patterns of behavior, some of these are patterns of art, of this maybe social patterns, may di different kinds of competencies that we see in problem solving. Yeah. Anyway, that's the that's what I think is going on. Matters a lot.

Michael Levin 48:57

Yeah, a couple of things. Well first of all, just before I forget, I wanna you know, you were talking about the ancient aliens and whatever. I mean, you know, people ask about all sorts of they say, well, you know, these woo things. Like there is nothing more woo than this than the fact. I mean, it's not a conjecture, it's a fact that there are that you start with logic, meaning I've got an empty set and I've got the successor of empty set, right? So a set of one element, and from that I can build up arithmetic. And before you know it, you find out that wow, E is exactly 2.78, blah, blah, blah, whatever it is. Well, you started with empty set and successor of empty set, and now you'd find out that it's actually 2.78. You don't have to do any physics to do it. You don't have to know a damn thing about the physical universe to get there. Can't do anything in the physical universe to change that fact. But that fact actually has all kinds of implications for how stuff works in physics, like the causation from that kind of world into changing things that happen, that is as woo as anything could possibly be. Once you've bought into that, and I don't see how you could get out of it, like that's really I think what's you know, un unassailable. Once you understand that, everything else is a pale sort of version after that. And that I think is a fundamental thing that people really need to get their, you know, get their heads around.

As far as, you know, my experience, first of all, I'm absolutely not going to claim that I'm the first person to have these ideas, right? A lot of these ideas have been said in various ways by you know, much smarter people back in the day. That's true.

Michael Levin 50:34

I think so. I yeah, yeah. So the only so I wanna say a couple of things. One is that I don't think I necessarily have said that no it's things that no one else has said, but I do think that my role is to bring these things into a logical framework that has practical implications. So yes, it may thousands of years ago they may have said some of the same stuff, and yet we've had three thousand years of cancer and birth defects and whatever, right?

So whatever it was that was said back then has not been integrated into our scientific understanding in a way that makes practical has practical implications for improving life. And so I think my role is to put some of that stuff together into a framework that is rigorous and connected to the scientific understanding of the world and not just like I don't want to add to the new age bookshelf. I want to actually make this to actually unearth the science that is required to understand these things.

Michael Levin 51:41

I would say a hundred percent. I don't have it to the extent of some what some of these people say. Well, well, I see the, you know, the poem running at me. I don't see that. But I absolutely do believe that one of the things that I've done from the time I was very young is work really hard to invite this these ideas and to sort of, you know, work at making myself a good antenna for this kind of stuff and the people around me and making an environment where if these ideas want to be out in the physical world in some sense, and I don't claim to understand how that works, then this is a then my lab is a great place to be. We are a lightning rod for this kind of stuff. And I work very hard to make that be true every day. And I certainly feel like I don't feel like I'm inventing any of this stuff. I feel like I'm discovering it and it comes through me. But you know, all I can do is the hard work of trying to get it out and test the things that are usable and you know, that can be turned into practical things that help and make life better. I love

Michael Levin 53:28

I mean what I would say is, and I'm you know, this is like I'm I don't have any hard answers here, but I think what we mean by fate, by when we use that word, I think what we mean are large scale patterns that require some degree of difficulty to get around. I mean, this is not a crazy thing. We know, like if you know, if you roll down a hill and especially if it's full of sand, you're gonna have a hard time getting out. There's a there's an attractor.

You had a chance to avoid it at one point, but you took a couple of steps down and now you've sort of slid down. So we know there are patterns in the world. And so if we believe in synchronicity and some of the things that you know Jung and Polly talked about, there are maybe large-scale patterns. Some of these patterns are trivial for us to navigate. Other patterns have their own pull the landscape of what's possible to you. And of course.

I think we make our own, we deform our own landscapes every day. Right. If you decide to go to, you know, if you do by changing your education regimen, your anger management therapy, you know, boot with meditation, you know, whatever it's going to be, you're deforming the landscape for your future self. And if there that there may well be, I don't have any idea, there may well be very large-scale patterns that are incredibly difficult to pull yourself out of. I up, you know, I tend to be optimistic and think that.

With the right amount of effort, probably most of those are not fully determined and can be avoided. But because we have such limited understanding of what's going on, it's really hard to avoid large attractors when you don't know what's going on. So I think that's what we mean by fate is I think we mean large scale patterns that without specific effort you're not gonna get out of.

Michael Levin 55:43

No, I've read a couple of his other books I haven't read.

Michael Levin 56:36

It does. Yeah. All right. So let's see, in the last two minutes, I know. First of all, I think emphasis on genes here is misplaced. I think things are much weirder and more interesting than that. Our two headed flatworms and one headed flatworms have exactly the same genome. It's not like even the inspiration that enables cells to build towards a particular an anatomical structure, even that's not in the genes, you know.

And if that's not in the genes, the this, you know, this other stuff, but there's no reason to think that it is. I think that the genes are a pretty good spec of the hardware. The genes tell you the parts that you get to have, the mo the proteins. I mean, we know we can read genomes now. We know what that they code for largely for proteins and a couple of other things. And it's just that just isn't what they do. So no, I don't think it's genetic at all. The other thing is I do think it's bi directional.

And it's bi-directional in two ways. One thing is that these patterns that ingress into the physical world, I don't believe they're static and unchanging. I mean, maybe the digits of E are. I don't know. But the vast majority of them, the ones that we care about, like us and other things, I think they are dynamic. I think they do change by their experience in the physical world. I have no idea what the dynamics up there are. But I don't believe at this point. Now I can't prove it yet. We're working on this. I we haven't proven that you can write information into that space, but I bet you can. I suspect. And we have experiments that we're doing where I actually think that you could do write information, maybe do compute in that space. I don't know.

Michael Levin 58:34

I think yeah, I even bubbles six lines of bubble sort can access this space. So

Michael Levin 58:41

We're just not good at detecting it. We're not good at looking at it. Even so this is my so this is my favorite one of my favorite examples of this kind of thing. So Alan Turing, who's a hero of mine, right? Just you know brilliant. In his paper on the Turing test, which everybody talks about the Turing tests, I don't I've never heard anybody talk about this part.

Towards the end of the paper on in on the Turing test, he says the following. It's actually probably not gonna work, he says, because the evidence, he says the evidence for telepathy is really good, and humans are gonna be able to tell when they're talking to another human, but they're but the computer definitely can't. And I was like, my god, man, like and I'm not I don't want to talk about whether you know telepathy is a thing or not, but he thought it was a thing. And the important part is that even as open-minded and as brilliant as he was.

And the whole paper is about artificial intelligence and right. He still, for some reason, said, Well, the humans can do it and the computer can't. Why? Why do you what do you know about this and this weird thing that you think the computer can't do it? It was it's an assumption. He was just assuming. And so this just reminds us all that we have these buried unstated axioms that we buy that we bind to from the earliest to I don't know, our earliest education, I guess, that doesn't allow us to see certain things.

And I was just like, why would you assume that? We have no idea. If it were real, like who knows how it works? Why the hell wouldn't a computer be able to do this? What do you know about the protonaceous, you know, squishy stuff that we're made of that for some reason enables you to do that and not this, not this computer? So I think we have to be incredibly careful and have a lot of humility about making any kinds of these pronouncements and we have to let the data lead. We have to do the experiments and we have to total up the free lunches and try to make sense of.

What that space of options is.

Michael Levin 1:00:35

Everything that we know about them right now goes through the language interface. We are obsessed with the language. The language is what we force them to do. When in the minimal algorithms that we study, the interesting ingressions are not the thing you force them to do. It's the thing that they do in the in between the thing you flee, like in the spaces you left for them, which are inevitable. And I don't know, you know, verbally what they can and can't do.

I think that's scratching the surface. I mean, you know, you your field is the first to tell us that whatever the patient says is like ten percent right at best of what's actually going on there. And I, you know, we are so obsessed with the language output of these things because we assume that we made it and it speaks and that's what the thing is. We're doing all kinds of experiments looking at their microbiome and at the, you know, all kinds of stuff that look potentially lives in the, you know, inside of them.

Michael Levin 1:01:43

We have no idea. It's the language output is the thing we force them to do and that may not at all be the interesting part of their cognition.

Michael Levin 1:02:01

Yeah, thanks very much. I appreciate it. Let's keep in touch. And if you decide to write that the write that review, I would love to I would love to see it.

Michael Levin 1:02:11

Cool. I appreciate it. Thank you so much.