Cross Word

Bridging the Gap: Exploring Consciousness, Physics, and AI with George Musser

• Michele McAloon

Share episode

Send us a text

What if the mysteries of consciousness and the universe are more connected than we ever imagined? Esteemed physicist and author George Musser joins me, Michelle McElhoun, on Crossword for an enlightening conversation that explores the fascinating intersection of physics, neuroscience, and artificial intelligence. Musser shares his insights into why physicists are increasingly drawn to these fields, suggesting they may hold the key to a unified understanding of our universe. Drawing from his book "Putting Ourselves Back in the Equation," Musser provides a compelling narrative on bridging the gap between mind and matter, urging listeners to reconsider the role of the observer in quantum physics and the humble perspective we occupy in the vast cosmos.

Our discussion ventures into the complexities of neural networks and their potential implications for understanding human intelligence and consciousness. Musser compares these networks to intricate webs of tiny switches, processing information in ways that challenge traditional computing models. We explore theories like integrated information theory, which propose that certain neural networks might theoretically possess consciousness due to their sophisticated feedback loops, and examine how these loops mirror the human brain's intricate systems. Musser emphasizes the engineering challenges these concepts present, while also highlighting their potential to revolutionize our understanding of both artificial and natural cognitive processes.

Curiosity and free will take center stage as we discuss their roles in human and machine learning. Reinforcement learning emerges as a fascinating example of how curiosity drives both humans and AI to persist through challenges without immediate rewards. Musser reflects on philosophical tensions between deterministic laws and the human experience of free will, drawing on his own life experiences to enrich the conversation. We wrap up by considering how AI might enhance our self-knowledge and deepen our understanding of human nature, inviting listeners to explore George Musser's work further and ponder the profound mysteries of consciousness.

Michele McAloon:

Hello, you're listening to Crossword, where cultural clues lead to the truth of the word, and my name is Michele McAloon your host, and we have a wonderful guest and, I can say, probably the first physicist I've ever had on the show, and I've asked him to speak real slow because I'm not that bright when it comes to physics, and he has assured me he will. He is a graduate of Cornell University, brown University. His name is George Musser. He is a contributing editor at Scientific American and Natos Magazines, a contributing writer at Quanta Magazines and the author of three books, and I urge everyone to go out and get these books Spooky, action at a Distance and the Complete Idiot's Guide to String Theory.

Michele McAloon:

He's a recipient of many writing awards and his book that we're going to talk about today Putting Ourselves Back in the Equation why Physicists are Studying Human Consciousness and AI to Unravel the Mysteries of the Universe and it is published by Farrar Strauss and Giroux out of New York and it is a fantastic book. He takes some very complex, some very, very complex material and really does write it in a level that I think most people can understand. So welcome to the show, george.

George Musser:

I'm honored and humbled to be here, and I hope I won't be the last physicist. I won't scare you off from going back into other ranks of our profession.

Michele McAloon:

No, I hope so, and actually I hope to maybe read a couple of other of your books and maybe we'll have invite you back for an interview. So let's start with an easy question why are and I'm saying easy in air quotes? Ok? Why are physicists beginning to take up subjects like neuroscience and artificial intelligence? What does a physicist have in that realm of business?

George Musser:

Yes, you might think there might be no connection, actually, that the world of matter that physics concerns itself with is disconnected from that of the mind, and so there's almost a schism in our culture and Western culture that goes back really to 17th century and goes back to the origins, or even earlier 16th century origins, of what we call science, natural philosophy.

George Musser:

At the time, and I think it was perhaps a tactical move, it was important to make sense of phenomena in the world by breaking things apart and taking the material content and considering it independently of the mental, and a huge progress was made understanding flights of cannonballs and why the moon orbits and chemical reactions and such like orbits and and chemical reactions and and such like. But I think science has progressed since then and is now reaching kind of a state where we have to go back and repair that schism. We have to reunite mind and matter and come up with a more coherent view, scientifically at least, also philosophically, of the universe, reality, the moment and there's very specific problems that I go into in the book and we can discuss, if you like, that kind of force the issue and force the physicists to reconsider the role of the observer, our own embeddedness in reality, the fact that we are mortal creatures who have a limited view of reality and have to be humble about it and take that into account in our theorizing.

Michele McAloon:

Let me ask you a question what is artificial intelligence? I mean that is thrown around so much AI, this AI, that AI is going to take over the world. But if you could yes, if you could kind of give us a short summation of artificial intelligence before we kind of launch into some of your other subjects in the book.

George Musser:

Absolutely. It's an uneasy question to ask. Everybody will have his or her, their own definition of what AI means, and sometimes it's like a marketing term, so you kind of do have to pick apart. What does it mean more intellectually as well, and I think of it as the the recreation in a machine system of human or probably mammalian or another animal intelligence. It's, I think, not different in kind from human intelligence and the intelligence of other creatures in the natural world. Still, even to this day, a dog has far more intelligence than any machine system. The flexibility of an animal to cope with the vagaries of life exceeds whatever a machine system can do.

George Musser:

So really, the question is intelligence? Intelligence can be, can take a natural or biological, shall we say, form, or it can take a what we might call an artificial form in a computer or maybe some other kind of automaton. And then the question is what is intelligence? And I would personally define intelligence as problem solving. It's the ability to take in the environment that we see and apply our memory, our judgment and solve a problem. Usually in life it's just to stay alive. It's to figure out how do we navigate a complicated and sometimes hostile world do we navigate a complicated and sometimes hostile world.

Michele McAloon:

Okay, so how does this relate to quantum physics? How does this relate to what you're going to talk about in this discipline of science called physics, and especially quantum physics? How does the human mind or intelligence have to do with this?

George Musser:

subject area and that you put your finger really on the kind of biggest. There are other questions, of course, that come up in this, in this kind of intersection of these different disciplines. And just as a footnote to that, I'm fascinated by interdisciplinarity and the boundaries between, like, domains of thought and domains of feeling. And I think we introduce those boundaries because it helps us but ultimately they're not in the world, they're things we need to sort of overcome. So there's really at heart no difference between physics and biology and chemistry and poetry and et cetera, but we introduce those boundaries conventionally and then we can come poetry and et cetera, but we introduce those boundaries conventionally and then we can come back and revisit. So that's kind of the overarching project of my work, not in just the book, only in this book, but also just throughout my career, is bridging these kinds of divides.

Michele McAloon:

So quantum physics yeah, now you're gonna talk real slow, okay, so what is quantum physics If?

George Musser:

I talk too fast and, by the way, I tend to be a fast talker, just even when I'm talking about anything. So just put your hand up and we will. We'll just, we'll back up and slow down and sometimes with quantum physics it takes a few rounds. Okay, it's, it's not actually that difficult, but it is unfamiliar, it is stretching us and therefore you and everybody have to be patient with yourself and then we'll just go around and around. We'll come to the topic, if necessary from different angles, and go from there. So quantum physics is the current understanding of the material world. It's often said incorrectly, so this is one of the first things we'll overcome Often said incorrectly to concern small particles, atoms, molecules, little things.

George Musser:

Small particles, atoms, molecules, little things. And that's indeed how it was created in the early years of the 20th century. It was a theory of the micro world, distinct from pre-quantum physics, or we say classical physics, which is the theory of the macro world, including humans, but also universe, galaxies. But what makes quantum physics or quantum mechanics, which is kind of a subset of quantum physics, so puzzling is that it doesn't have such a limit. It actually applies to everything. It applies to atoms and to us, and to galaxies and to the universe, and there actually are phenomena in cosmology on the vastest scales that we have access to as humans, that are quantum. So it's everywhere. But often we talk about it and I will give examples in terms of particles, but always in the background of our thinking is that it isn't just particles. You can't sequester it in the kind of microscopic realm. It's something that we have to deal with throughout the world, throughout the material world.

Michele McAloon:

So it's the building blocks for our material existence, whether it's atoms or neutrons, protons, all of that kind of stuff, and it can be small and as you talk about in cosmology, your chapter on cosmology, it can also be infinite too, exactly.

George Musser:

So it. I don't know how to. I can't think offhand of a good analogy but it's like a toolbox and you can take out your hammer, screwdriver, et cetera and apply to a nail screw, but you can also apply to hit the side of something to get it working. I mean, you can take your tools and apply them in multiple realms, and so it is with quantum physics. Typically, we apply it to particles, but we can also apply it to these other systems. We were forced, however, to apply to particles because they really exhibit some of these peculiar and I'll go into a second what I mean qualities, behaviors, more than a human or a larger object would.

George Musser:

There's a whole long history and it's quite quite entangled, so to speak. Actually, there's a double entendre there, but the essential kind of tension or thing we have to wrap our heads around is that everything in the world, all objects, all material objects, can behave in two seemingly contradictory ways. They can behave as particles, and by particle here I don't mean necessarily a little thing. What I mean by particle is a localized clump of of matter or energy. It's something that has a position, it has properties such as color or charge or mass. It's a chunk, it's a thing. I guess, robbie, that's how we would just say it in regular speech.

George Musser:

Or we can think of the components in the material world as waves, and a wave is something that's spread out, it's a distributed phenomenon. It's not in a particular position sound waves and they're going into the microphone, but they're also going into the house, going to the yard, they're going out into the atmosphere. So it's not just confined to one place, it's distributed. And in fact all the properties of waves wavelength, for example, frequency aren't properties of a particular any point, they're properties to describe this distributed system. So anyway, quantum physics says or is rooted in the observation. It's not really, it doesn't say it, it's just it's our way to make sense of the observation that objects in the world can behave as waves, distributed, or particles clumped.

Michele McAloon:

Actually that's very clear. That is not hard to grasp at all. We made some progress, okay, but and your book really goes into how these two things quantum physics, wave, collapse and we'll get back to that one but how that goes into understanding human intelligence through artificial intelligence, and that really kind of I think is the basis of your book that we are material beings in a material world and that we can understand ourselves materially, the same way as we use these machines to try to replicate human thinking, thinking we. Again it comes back to quantum physics. So I have a question for you. So you have a couple things in here. You talk about what is a neural network, and I think of a neural network as that gray matter in my head. But scientists have replicated, I think, since the 1950s, these neural networks. Tell us specifically what a neural network is?

George Musser:

These neural networks Tell us, systems that run chat, gpt, for example, or image recognition systems, current technology for language translation goes through a neural network concept that sometimes we just read in the papers neural network, artificial neural network, chat, gpt, whatever and we kind of don't stop to think about how just wondrous that concept is. So our brains are some billions, tens of billions of neurons, cells, and there's other, actually other kinds of cells in there in the brain that are actually quite interesting in their own right, but I'll just focus on the neurons. The concept of the neuron as a distinct cell type goes back to the mid-1800s really, and it was a big revelation within biology, the emerging field of neurobiology, of how the brain works. It wasn't really understood what it was before that, and so actually their neural network goes back also to the mid 1800s because the initial kind of impressions and microscopic images of brain tissue showed these neurons, nerves connected them. Nerves are just actually parts of neurons and they link together in a vast kind of fisherman's net, or I think of it like an airline route map kind of, or train network, subway map kind of thing. So there are points like subway stations that have lines like train lines between them and each neuron itself is like a little computer.

George Musser:

It's actually to this day not fully understood all a single neuron can do. It's actually extremely complicated but normally is modeled or simplified as a switch. A neuron is a teeny switch, it can be on, it can be off. It turns on based on the kinds of inputs it gets, turns off based on other inputs that it gets, and it's connected through its wiring to actually thousands of other neurons. And what's cool about it is it's a pretty simple kind of setup. Right, you've got a switch and wires. You can imagine I actually again behind my computer screen here you can't see it is.

George Musser:

I made my own neural network from relays which are type of mechanical, electromechanical switch and wires, literally wires, and by wiring the, the neurons, these switches, in certain ways you can have them achieve a, produce an output for a given input, so they can do computation. And it's a different model of computation than we currently think of in a computer. A computer has what we call a CPU, a central processing unit. It's a central element. It basically takes instructions, data, follows the instructions to do something to the data, like add, subtract, actually move the data from one place to another. That's actually far the biggest component of computation is actually just moving data from one place to the other, and it does it in a serial fashion. It does it one after the other, a sequence, just like you and and I would, if we're working through our errands list, would you know, get the soap and the aisle, then we go over to the produce aisle and whatever we have a list of things, and so does a normal computer, whereas a neural network is a distributed computer.

George Musser:

each neuron is extremely simple, just a switch, as I said, but collectively they act.

George Musser:

In the case of the brain, billions of them act together to perform a computation, and there's no one place that the brain billions of them act together to perform a computation, and there's no one place that the computation is being. There's no one place where the addition is being done or a movement of data is being done. It's done over the whole thing. If it stores a memory, if it takes in information and retains it in some way, it's not retained in any one place, it's retained over the whole thing. It's a spread out notion of computation that turns out to be exceptionally powerful. Certainly, it's a huge number of kinds of operations and I think, with all the progress that's being made in creating machines, versions of this, in other words, artificial neural networks that run on computers, run on your phone, now we're actually understanding ourselves better. I've had several articles that have pointed this out that now we have a better concept of what human intelligence is. Human intelligence is a distributed computation. It happens over vast parts of our cortex regions of the brain.

Michele McAloon:

Okay, I did something that you recommended or in the book that you talked about. I downloaded a little neural network app and I've been sort of playing with it and I think it's based on Cobra and it's a little bit like monkeys play with buttons. But you know, I understand it a little bit better. Physicists, which I understood from your book. They're pursuing kind of a united understanding. They want to try to find a singular explanation of the universe where all of this is united together. All this material world is united together.

George Musser:

That's correct right, I would say that's the overarching goal of physics, and I would actually define physics. What does it mean to be a physicist? What does it mean to do physics, right, okay, means to find the simplicity, and the complexity, is to find the unity and diversity, and there's other domains of the world that rebel in diversity, and that's wonderful. But physics, what it does, is it looks at all the things happening in the world and tries to find a unifying principle to them. So yes, by its very nature, physics seeks unifying explanations, and we're at the point in the development, historically, of the subject where we're ready to do the whole thing next. That's like the next leap is going to be not just unifying this part or that part, but unifying everything that we see.

Michele McAloon:

Okay For the human experience. Our unifying experience, I would have to say, is probably consciousness, to be conscious, right? How does what we're talking about? We're talking about neural networks. Ultimately, we're talking about machines or something artificial that will replicate our intelligence. So let's start with what is consciousness.

George Musser:

Distinct from intelligence, but an open question, one that I've been exploring. I don't really get too much in the book, actually, but one thing that's relevant is how is consciousness related to intelligence? Are they necessarily related? Can you have an intelligent being that's not conscious or a conscious being that's not intelligent? Conscious being, that's not intelligence, but conceptually they do come apart. So intelligence, as I said, is like a problem solving ability that, given inputs, it tries to make a prediction or an output or what have you, and that can be in many forms. We underrate the types of intelligence.

George Musser:

One form of intelligence is visual intelligence is looking out in the world and seeing it as composed of objects that have colors, textures et cetera. That's a form of intelligence. Because it's a problem we take retina impressions, light into our eyes, and the problem is how do we figure out what created those light impressions? So arguably the greatest intelligence on our brain is that of the visual and the other perceptual courtesies. So in consciousness, however, is the, at least as I'm defining it. There's, again, multiple definitions. I I focus on what's called phenomenal consciousness against a philosophical term, but what it refers to is the having of experiences against a philosophical term, but what it refers to is the having of experiences, that I as a human, don't just take data in, I don't just form impressions, for instance, as I said, of the visual world or the sound that I take in, but I also, on top of that, have an experience of it.

George Musser:

I don't just take the red in, I feel red, I feel the redness when I see something red. In fact, a little recording button on my screen, now that you started, is red, so it conjures up oh, it's like a fire engine, or like a fire, or it's like the sunset. I have all these associations with red. I feel the red. It's not just that I see, it's red, I feel, feel it. That's the conscious component. And again, some people, some philosophers, neuroscientists, think that that's part of being intelligent, that to process that button that's red on my screen, I have to have these associations. But others think no, it's kind of sitting on top, not necessary to it. So what comes up in physics is that there are points in quantum physics the act of observation seems to require a conscious perception. It's not just a mechanical or even intelligent registration of information, it's also that there's an experience being had there, and that's very peculiar and that's the kind of puzzle we're trying to solve and that's the kind of puzzle we're trying to solve.

Michele McAloon:

And now you do a really good job of showing how the observer, whether he's in superposition, as you call it, or observing, is actually an integral part of the physical world, of actually what we know of reality, of what we know of physics, what we know of science. It is the observer in the world that sees this. Again, how do we take this back to machine language? How do we take this back to artificial intelligence?

George Musser:

now, yeah, so the connection to AI is a few connections. To take it back back, as you say, One is the kind of understanding connection. How do we understand our own act of problem-solving intelligence and experiencing consciousness? And to achieve understanding really anything that we do in our lives we really have to understand how to also build it. We have to be able to create.

George Musser:

True understanding usually entails that one could at least imagine. Even if you can't actually do it yourself, you could at least imagine how is something created, you can imagine how it's put together and the parts that are required. So that's one, I think, arguably for me intellectually the most interesting part about AI. It's not typing the questions in chat gbt, it's just taking what chat, gbt and the other language systems etc are telling us about our own consciousness and that can then be in turn fed into this observer capacity and in physics. So for me that that's really the main thing. And there are ways to understand consciousness, ways to understand intelligence in this network, newer network way, and once we kind of have that understanding, we can then bring it over to physics and perhaps begin to solve some of these puzzles.

Michele McAloon:

Because of the power of observation and how we are integrated. We're so integrated into that we're not only the problem, we're also the solution, as an observer, because if we can't observe it, then we can't measure it. Is that correct?

George Musser:

Yeah, I would actually use the terms in this context. Observation and measurement is interchangeable. When I observe something, I kind of implicitly also measuring it. I kind of if I see the paper on my desk, I don't actually go out and measure it like with a measuring tape necessarily, I just note the papers on this, but I also have a general idea of how big that piece of paper is. So I've performed a type of measurement. And going the other way, a measurement is the process of going into the world and taking reading of it, for example, and that also entails an observation.

George Musser:

And, as you say, we're embedded in the world. So our embeddedness, the limitations that we have, or the mere fact that we have to perform interactions to do these measurements and observations, condition the types of measurements and observations that we make or don't have a god's eye view. That's reserved for the divine. We have a temporal or a mortal view and we can't fool ourselves and often we hide that fact. We temporarily, provisionally fool ourselves and imagine we have a god's eye view and that helps in some cases. Sometimes we can stand outside the system that we're studying, but fundamentally we don't fundamentally have this interior view, this embedded view of the universe and have to take that into account. For instance, when I observe a quantum particle or observe any particle, really I don't stand outside a particle and imagine it in some platonic realm. I am doing something to the particle, I'm touching it, I'm causing it to veer in one direction or another that actually has.

George Musser:

Some of these measurements are done with a magnet, for example. We put the magnet in its place and the particle goes left. We think it has one maybe negative charge, let's say, and it goes right in as positive charge or spin or some other property that we're measuring. So I've done something to the particle in order to extract information from it. It's not a passive operation, it's an active operation. And when you and I and anyone look out of the world, when we look at it, we're not really interacting, we're not really affecting it that much. Like the book on my desk, I look at it. Very little is happening in the book, but there is something discernible happening in the book if you kind of look at it at a deeper level. And what's peculiar about quantum physics is we're always affecting what we interact with. There's no kind of even approximately passive measurement that we make of the world. Every measurement we make has a discernible and sometimes defining role in the world, and that the whole puzzle we're trying to solve is why and how.

Michele McAloon:

You talk about. There's a field called relational physics, right? This is what relational physics looks at is our interaction with the quantum world. Is that correct?

George Musser:

Yeah. So there's a deep philosophical problem that I call and other people have called but I adopt their term the hard problem of matter. Yes, there are a number of hard problems and that's kind of almost like a term now the hard problem of mind, the hard problem of life. There's another hard problem and I in this case focus on the hard problem of matter. So the hard problem of matter is what is matter? What is it exactly? Matter is?

George Musser:

Physics describes electrons, protons, and you name it and it does. But it tells you what the matter does. Physics tells you that if you have an electric field, the path of a charged particle is diverted, something like that, our magnetic field. It might cause or be caused by electric current. So it's describing happenings, behaviors, but it didn't really tell you what the thing causing the behavior is. It doesn't. There's no real.

George Musser:

We don't, in other words, talk about the intrinsic nature of the world. So one answer is one way to understand that is is that the world has no intrinsic qualities to it. It's all about what the objects in the world are doing. And this is where we get into this idea of relational or structural realism is another philosophical term for it. We get into this idea that the world is relations and there's really nothing being related, it's just all relations all the way down. So electron is a bundle of of relations among objects and it. You might intuitively a lot of people intuitively have trouble with that.

George Musser:

Okay, this is the idea and we can dissect it later if you, if you like, and a lot of the kind of puzzles in quantum physics and these other domains that I talk about can be kind of brought back to this relational view that everything in the world is relation.

George Musser:

So when you do a measurement you're establishing a relation with the thing that you're measuring, be it this pen or electron et cetera, and the measurement that you take is not a flow of data, like the pencil is four or five inches long, but it's a relation that you've established between that. So the four inches of the length of the pen, for example, really describes something about you and the pen together. And Carl Rovelli is a physicist I greatly admire and he's really been working through this relational view. It has its problems. It does lead to some. I don't fully accept it myself, or at least I don't think it's been developed enough that we can fully accept it. But it is a very powerful idea and a lot of the puzzles I talk about in the book can all be kind of put under that umbrella of relations. Understand the world as a set of relations.

Michele McAloon:

Okay. So when we build neural networks you have something called integrated information technology and we'll talk a little bit about that that sort of emulate the human brain. Are we capable of doing that? In what you just said, when we build these things, are we basing it on relationships, on measurements for language programs? And if we're doing this on relationships, then we're actually doing it on experience, because relationship entails an experience. So, as I talk to you, there's an experience, it's a relationship, but there's also an experience in that relationship. And I remember what I said to you five minutes ago. Right, I will remember. I'm thinking about questions that I'm going to ask you that hopefully sound smart in five minutes. Right, they do so. Is a machine capable of doing this? Are we at that point of science?

George Musser:

Yeah, there's a lot packed into your question.

Michele McAloon:

I'm sorry.

George Musser:

No, that's fine. So let me just prize out the relations just for a moment. We'll put it on the side, because I think when I talk about relations in this kind of discussion I'm really talking very fundamental picture here. We don't always have to be using that picture. This is the thing about science that works on different levels. So we don't have to adopt a relational, deep, foundational level of discussion to think about things above it. I can perfectly adequately talk about baseballs and pens and living things without breaking them down into the relations that they embody. That's kind of a different level of explanation. The relational is a deeper one and the idea of a thing is the higher one in this particular case. But there's other levels of explanation that we can find. Actually, frankly, I think this is one of the most profound aspects of the world that we seldom reflect upon is its level nature, the fact it can be conceived of as multiple levels of description. So let's put relations on the side and we can come back to that later.

George Musser:

Integrated information theory is an utterly fascinating view of consciousness and it does tie into or can be applied to, perhaps tying into quantum physics and our understanding of the puzzles of the measurement process in quantum physics. And what integrated information theory does is it describes a conscious being humans, but really any conscious being. That's actually what's cool about it is it can be applied to really anything, anyone, anything as a network and, again, much like any kind of. It's really is a neural network. It's a bunch of little switches, neurons, that are wired together by nerves, or however you conceive of neural network wired together by wires. Actually, in case of iot, it says that there's a particular quality, a particular type of wiring that leads to conscious. There's not any neural network. In fact, chat, gpt does not, according to IIT, have the type of wiring needed to be conscious. Most neural networks do not. Neural network that performs language, translation actually maybe language, but not really. Most neural networks that we talk about in technology, that you hear about in the papers, et cetera, are not capable of this.

George Musser:

What IAT says is that there has to be loops in the system. It has to be what we call recurrence. Information has to go around and around and the output has to go back to the input or somehow modulate the input, and in these loops you get new dynamics that, according to that theory, can be identified as consciousness. In particular, the dynamic you get is of overall system coherence going around and around in a loop. It goes out your mouth and into your ear. Then the system has like an overall unity to it. If you start breaking it apart, you lose something. If you cut the loop, you've lost something, whereas if you cut out part of chat DBT, yeah, it'll be degraded a little bit, but you're not really fundamentally losing operation of chat GBT or the kind of neural network that runs chat GBT, whereas with a recurrent or kind of a feedback network, as we call it, there's something very fundamental that is lost when you start slicing and dicing it. And IAT, this integrated information theory, so-called kind of has a whole quantitative analysis of the loop structure within networks, including perhaps our brains.

George Musser:

Our brains are thick with loops. That's actually what makes them interesting. What makes them extremely hard to understand scientifically is that there's a lot of feedback in the brain. So we experience this in daily life. One example the psychologists and the neuroscientists love are just visual illusions, like when we see, you know, the see in the image, colors. What we see, and what we see affects what we expect to see, and those go in a type of a loop and sometimes in these puzzles you kind of flip back and forth. You see the vase for like three seconds and it goes over to the people or whatever. I can't remember how that illusion works. But it goes back, flip back and forth three seconds and goes over to the people or whatever. I can't remember how that illusion works, but it goes back, flip back and forth three seconds and goes, and those kinds of locking in for a few seconds of particular interpretation of an image is related to this loop structure that we have in our brains and the psychology that sits on top of that.

Michele McAloon:

Have we been able to create anything that did you know? And I kind of liked your, your your explanation there, anything that did you know? And I kind of liked your explanation where you said you know, a neural network is more of a linear, where this integrated information is more of like a flood into the brain. It's water seeking its own level and I love that explanation. But have we been able to create anything like that?

George Musser:

Yes. So, yes, the answer is yes and actually a really interesting line of work, partly inspired by integrated information theory, which has been very influential in this area. People, you know engineers have sought to build systems that have this type of recurrence. In fact, a lot of systems, including the older models of older AI systems for language, did also have this recurrence in them. It turns out there's technical reasons.

George Musser:

It's very hard to build large versions of this. They're hard to train. So in order to train a chat GPT, which takes, you know, vast sums of the entire internet, they had to kind of basically straighten it out. They had to make it from a feed forward excuse me, feedback network into a feed forward network. A feed forward network is like a one-way street rather than a two-way street, but always in the background here is that a recurrent network might have this kind of capability. That doesn't necessarily, by the way, you have to accept the premises of the theory and you have to build systems. But yes, one possibility for building conscious machines either deliberately, because we're interested in building conscious machines, for either deliberately because we're interested in building conscious machines, for whatever reason, or accidentally. It's quite possible that we'll build a conscious machine and not know it. That would involve some kind of recurrence or looping feedback. But yes, it's completely within our technology to do that.

Michele McAloon:

One of the things and you didn't touch upon this, but it's a question when I was reading your book and I was thinking about, and actually iteless curiosity that human beings have from birth to death, of why, you know where, from how do we make sense of all of this? Do you think machines could ever be curious? Because I think that's really what makes us in so many ways human, in so many ways it defines our humanity, of whatever you believe, and curiosity is the to physicists and scientists in different fields around the world. I mean, you're talking to people in India and Africa and the US and Canada and France and Switzerland, and you're talking to all these wonderful, these great minds that are actually, I think, unified by this crazy human thing called curiosity that is insatiable, that we can't get enough of.

George Musser:

That's beautifully put. I really I do think that that is one of the defining features of humanity. I think actually we have many. But certainly curiosity, would you know, is an animating force in my own life and I think we see that culturally and broadly as well. Can that be programmed into a machine? Absolutely, I think we humans that culturally and broadly as well. Can that be programmed in a machine? Absolutely, I think we humans are a type of machine. Ultimately this is kind of also a background assumption we're not a machine in the sense of maybe a purpose-built system, but we do have mechanistic qualities and probably a lot of our thought processes can be understood mechanistically as the incredibly complicated action of billions of neurons. That has not been replicated yet in any machine system, even something of the order of a GPT-4. So I do think it's conceivable, but GPT as it stands, despite its size and complexity, does not have it. I didn't get into it in the book. I have written elsewhere about curiosity in machines and there actually is a kind of subdiscipline that thinks about these issues and it comes often in an area called reinforcement learning. And reinforcement learning is.

George Musser:

There's different types of learning theory here, different types of ways a machine can learn. One, the one that we often talk about I talk about in the book, but you know comes up often is discussion of neural networks is the machine can be presented with images of dogs and cats. That's usually the example we use because it's fun to use that and also actually that's really what they do. They use dogs and cats and one is labeled a dog, one's labeled a cat and the machine learns, by being given examples, how to distinguish those two categories. But there's other types of learning where the machine is given a reward for a solution. So learn to play chess, go, and it isn't really given a lot more direction than that. But if it learns to play chess or go, or learn to play, do something else I've seen simulations of little cars driving on a landscape, for instance or learn to drive. In that case, at least in that simplified case, at least in that simplified case and if it achieves that it gets a reward. And it's quite a general way of learning, because it doesn't kind of presume that it has to be given examples to.

George Musser:

That can manufacture its own examples, it can find whatever way it wants to get that reward, but it has to navigate through the kind of, if you imagine, a space of possible solutions, an abstract space of solutions to get to that reward, like a rat going through a maze. How does it do that? How does a rat? In fact, how does a rat figure out to go through a maze? Well, a rat goes down this passage it didn't work, goes to the other passage, eventually, finds its way out of the maze and gets its reward. And so a rat has to exhibit, has to kind of push itself through difficult tasks that don't seem very rewarding. I've gone down on, the rat says to itself. I've gone down on another passage in this maze Still no reward. And you might expect, rat just gives up and just lies there. I mean, why would the rat continue to go through that maze?

George Musser:

Because it's driven by something, and often we can conceptualize that, in the case of rat and certainly in the case of a human, as curiosity. So curiosity is what propels us through life when life doesn't have any other reward. Right, often, maybe that's the majority of our experience we are just toughing it out. We are going through this, this world, and we're trying to make sense of it, and there's not a lot of feedback. We're getting what is pushing us forward Curiosity. It's what we have in the absence of some kind of external reward.

George Musser:

It was always just a reward, and this is sometimes I think this is drilled out of in school. We get too many rewards, right right. We're not forced to develop an internal feeling within our gut that we're just going to go for it. And you read these biographies of scientists and often they have difficult childhoods that didn't do well in school. They didn't get that positive feedback, in fact they got negative reward in a way, and they had to have that fire in their belly to continue going. And eventually they get a Nobel Prize 30 years later. But that's a long way to be. You don't just see a Nobel Prize in distance and go for it, you have to just push yourself through, just sheer bloody mindedness. And that comes, I think, from our sense of curiosity.

Michele McAloon:

And you know what? You're a very brave scientist because you touch upon free will and I actually like your conclusion to it. You know, are we predetermined? Are we free to make the choices off of our experience and of experiences that might have been determined, but we have that freedom to choose what to do next. And I ultimately, I mean personally that's have been determined, but we have that freedom to choose what to do next. And ultimately, I mean personally, I think we have radical freedom and we don't handle it well. A lot of times we don't. We would like to be predetermined, but we're not, and I thought you handled it very well in the book.

George Musser:

You're very kind to say that, because I'm actually thinking of taking out some of those ideas on free will and writing them up separately. So I've been doing some reading on free will lately and there's a lot of discussion about that. I should say you and I haven't talked about this before that I actually have a Catholic upbringing myself. I went to Catholic school for a couple of years and CCD my whole childhood and it was free will. That was always a huge issue for me and I remember that from my childhood.

George Musser:

Because if God knows everything, what freedom do we have? In what meaningful sense do we have freedom? And that puzzle, that dilemma, is very much mirrored in science. So if the laws of physics are thought to be deterministic, that everything that happens is kind of built in to the present moment, there's no freedom at that level. Yet we of built in to the present moment, there's no freedom at that level. Yet we of course have the experience of freedom.

George Musser:

And I think we do have genuine freedom, even if the laws of physics are deterministic, because I understand freedom as the ability to make choices based on our life experiences and based on what we see. It doesn't mean we stand outside the world, we don't have a kind of our own personal godlike power to stand outside the causal order of things. We're in it. But nonetheless we are free when we act on the basis of our desires and our deliberations and we come out with some kind of product of that. It flows through us.

George Musser:

So it free will for me is. It's a project of self-definition. It's what does it mean to be us? What does it mean to be you, me, any individual in the world? It doesn't mean that we're not influenced by stuff. I'm influenced by my childhood, by my parents, by my culture, by the sunlight coming in the window. Of course I have all these influences. That doesn't take away my own autonomy that those influences flow through me and then back out into the world again, and to me that's freedom. Not everyone accepts that view. By the way, it goes under the rubric of compatibilism. That's the philosophical term here. I'm a compatibilist, but not everyone is. Some people I greatly respect disagree with that, but that's my view.

Michele McAloon:

Not everyone is, and some people I greatly respect disagree with that, but that's my view. I do, and you free will, and you tie that into actually personal identity and I just I think it's absolutely right. I could talk to you for hours. This book is absolutely fascinating and it is written in such a way that it is accessible to anybody who wants to take the time to read it. It's not a fast read and there's, you know, like the what was it? It was quantum collapse or wave collapse.

Michele McAloon:

I mean, I had to read that like three or four times. I had to look on the internet to find out what collapse is, to actually understand it, because I don't come from a science background. But you write it in a way that is not scary. You write in a way that actually explains and ultimately, I think it's a book about hope, because as we move forward with artificial intelligence, it's not going to replace us as human beings. It might replace some grunt work. There's something different about human beings, but there's also something that we can find out about our human selves through artificial intelligence, through looking at the mind, and we should have the confidence and the fortitude to continue with these studies so that we understand ourselves better, and you know what I think that's the ultimate human experience is to understand ourselves better, and that's what we got here.

George Musser:

So that's really eloquently put and we often get knowledge in science. But I think self-knowledge is part of that. To understand why do I do what I do or how do the people around me behave and what am I doing on this earth is hugely important. I frame my entire life around that. I think most scientists do. I think most people do, even if they don't articulate it that way.

Michele McAloon:

Maybe you and I have an article about the nexus of science and theology. You know that, and free will, maybe we might, maybe so, folks, this is putting ourselves back in the equation why physicists are Consciousness and AI to Unravel the Mysteries of the Universe. It's by George Musser and it's Farrar Strauss and Giroux of New York Cannot recommend this book enough. It would be a great Christmas book and I believe. Where can people find out more about you?

George Musser:

I have a website, georgemussercom. They can go there. They can follow me on whatever social network they have. I'm mostly using Blue Sky these days, so just look me up there. Probably is the best.

Michele McAloon:

Okay, great, all right, george, thank you so much. We really appreciate you taking your time out of your very busy schedule to talk to us, and we look forward to speaking to you some more actually in the future.

George Musser:

It's my pleasure, thanks.