River

You're listening to The Emergent Podcast, Episode 10. Ten episodes in, Justin and Nick have circled this question from many angles. The alignment problem framed as a moral problem. Machine ethics. What it means to relate to a thing that talks back. Today, they meet it head-on. Dr. James Bigelow, a neuroscientist who spent his PhD watching memory unfold in macaque brains, joins them to ask whether the spark that lights up the inside of your skull right now can ever, even in principle, be detected from the outside, or whether the inside is all there is, and the rest of us are simply taking each other's word for it. Here's Justin.

Justin

Welcome one and all to the Emergent Podcast. We're here on episode 10, and we've got the neuroscience of consciousness on the docket for today. I'm joined as always by Nick Bagley. Nick. Hey, everyone. Good to hear you. Justin Harnish here. And we have a special guest on the phone with us, Dr. James Bigelow. James, really a pleasure to have you on the

James

podcast. Yeah, same. I've been looking forward to this. This is something I like to think about and talk about and looking forward to our discussion. Outstanding. So we want to jump

Justin

right in with your expertise, James, and really talk about the current state of the neuroscience of consciousness and where is the current research at what is keeping consciousness researchers up at night and maybe even before that how does a neuroscientist frame or define consciousness

James

yeah well let's jump right in so a little bit about my background i guess would be irrelevant I'm not a consciousness researcher. I think it's interesting, and I pay attention to articles when I see them. I actually got into neuroscience. Basically, my PhD was studying the neuroscience of learning and memory. We used, in my PhD, we used macaque monkeys and studied brain regions during a memory task performance. And we tried to kind of use the activity patterns to decide what the time course of activation among different brain regions was for supporting short-term memory. This is something that I actually think is relevant to the study of neuroscience of consciousness. The neuroscience of learning and memory, I think, is a much bigger field with a lot more mature kind of state of the science relative to neuroscience of consciousness, which is really populated by a very small handful of researchers. it's also just not been around as long as like a formal scientific study compared to learning and memory and I think there's some overlap you know I think the neuroscience of consciousness can learn a lot from the neuroscience of learning and memory some of the things that have taken decades to establish it's probably useful to I know you have both talked about the kind of what is consciousness on the podcast before. And it's probably really useful to talk about kind of what are the scientific approaches to consciousness and then what is still kind of falls under the domain of philosophers and theory. And I think that is nicely encapsulated by the easy versus hard problem of consciousness, right? So just to recap, this is a distinction made by David Chalmers, he's a consciousness philosopher, and he talks about the easy problem of consciousness is essentially determining what brain structures and what types of brain activity are all necessary for consciousness, conscious percepts to happen in people. But the hard problem is, given that all of those structures are intact and all of those activity patterns are happening, why does consciousness then happen at all? In other words, why does qualia exist? So qualia is like the redness of red, the painfulness of pain, and so on. It's the experience that accompanies perception that's accessible exclusively to the owner of the brain, right? So right now during the podcast, we're both having auditory experiences, right? And we can think about that from an information processing point of view. your neural networks can take the sound pressure waveform as input and parse the different spectrotemporal features and process what it is that I'm trying to communicate. But why do you then also have an accompanying experience of the tone of my voice and the loudness of my voice. So the hard problem is the hard problem, easy problem. Distinction has been something that's been really useful for decades in kind of like distinguishing what can we do to study consciousness in the brain versus what is accessible only via our direct experience. In other words, what's measurable and what's not. Are we calibrated by the way? Can I get a temperature check here? Are we on the same page?

River

Do you have anything to qualify or follow up on? Quick orientation. David Chalmers' easy problem isn't easy. It's tractable. Map the circuits. Find the activity. Correlate it with what people report seeing or feeling. Decades of careful empirical work. The hard problem is the question that survives even after the easy problem is solved. Why is any of this accompanied by experience at all? Why does processing wavelengths produce the redness of red rather than nothing? You could, in principle, build a being that does everything a human does and lacks an interior altogether. Chalmers calls it the philosophical zombie. The fact that we can coherently imagine such a thing is what makes the hard problem hard.

Justin

No, I think that that's exactly right and makes a clear distinction is that really the heart problem is, you know, given even a complete understanding of the neurocorrelates of consciousness, how is that responsible for this immeasurable experiential nature of our you know conscious makeup and so I think that that describes it really well and obviously those you know easy problems are also easy and have required years of research to plug in the components of the the contents of consciousness that folks oftentimes associate with consciousness itself. But there's this construct, right, that is what it is like to experience. And given greater intelligence, greater memory systems, a narrative structure, a mortality, a practice like meditation experimentation with psychedelics whatever the nature of those things are that alter and expand those contents is also interacting in some interesting way with that hard problem first person experiential conscious yeah yeah and before you jump in james i just

Nick

You know, Justin said the word how, so the engineer in me kind of wanted to jump in here and provide some other thoughts. One is that as we think about the potential distinctions out there, you know, as an engineer, I want to start breaking it down into systems and start providing these really strong distinctions that we can use to play and really start forming up. What actually can we solve for, right? And one of the areas that a lot of groups are really trying to focus on is looking at the level of consciousness versus the contents of consciousness specifically. And the other one that I think really stands out is that neural correlates of consciousness and getting into that versus like the prerequisites of consciousness and really the consequences of consciousness. So just curious, if we kind of frame it in those areas, what do you see or what are some of the areas that tease up for you?

James

So you said the contents versus the quality of consciousness.

Nick

Sorry, can you repeat that last bit? Yeah, really level of consciousness versus contents of consciousness.

James

I see. So like maybe some other organisms have kind of a rudimentary form of consciousness that there's something there, but it's like a quantitatively different level. Is that kind of what you have in mind?

Nick

Exactly. And I can go into more specifics, but I would be thinking about things like, what is the importance of this consciousness, the specific level being recurrent? You know, I've had this thought before, this thing is coming back again. Different ways that we can actually report it or have it be evidence-based. Thinking about levels from, you know, is it temporally extended? Is this something that goes over time, over generations, over other means? And how that kind of level of consciousness really kind of expands as well. Self-modeling would be another interesting level, right? Do I actually have a representation of self?

James

Well, let's just keep talking and kind of see where this goes. One thing that kind of can complicate discussions of consciousness is that sometimes different people have in mind different things when they mean consciousness. So maybe let's go back to the hard and easy problem. One thing we can talk about is like machine vision systems that can easily classify colors, right? Or shapes or letters. And so it's trivial to have the output of like a computer vision system tell you whether something is red or blue or green based on a photograph of something. And I can do that too. You know, if you show me a few different options, I can easily tell you whether it's red or blue or so on. um so the i think what what i have in mind when i say consciousness is everything that happens beyond and or like in correlation with the perception and the behavior so we can have systems that can perceive and respond to inputs in different ways but that are not necessarily conscious and i guess one of the constructs that's useful for talking about this is the philosophical zombie, right? You've probably wrote about this idea where you imagine being that's human in every way, except for that it lacks consciousness. So it can talk to you, it can respond, but there's no conscious correlate to anything that it's experiencing or saying or doing, right? And so that one difference kind of, I think, defines well what I have in mind when I say the word consciousness is the qualia. It's the experience that accompanies all of our perceptions and actions that we're aware of. Now, strictly speaking, the only evidence we have of consciousness is our own consciousness itself, right? You two that I'm talking to, for all I know, could be extremely sophisticated zombies that are talking to me and have the ability to interact with me, but do not actually have conscious experience yourselves because I am not observing your conscious experience, right? So that gets into Descartes' notion of radical skepticism. The only thing that I can be certain about is my own consciousness. Now, given that technicality aside, I think most people agree that all humans are conscious, but it kind of points out the devilish fact that we have no way of measuring consciousness. And just to double down on the hardness of the hard problem, you know, it might be an impossible problem. Not everybody knows that Francis Creek, after working on DNA for decades and winning the Nobel Prize, then went on to spend kind of the rest of his life working on neuroscience and specifically the neuroscience of consciousness. He wrote a book-length kind of summary on his views and then a couple of review papers, including one that was written right at the end of his life. In the introduction of that paper, he basically says, you know, nobody has made any progress on explaining why the experience of the redness of red arises from the actions of neurons of the brain. and it appears to be a dead end for empirical science. And so given that the hard problem of consciousness is so hard as to maybe potentially be intractable, he went on to discuss the easy problem, which is defining the neural correlates of consciousness. So I think the easy problem is easy by comparison because it's at least in theory, a tractable problem that we can develop, but it's also extremely difficult. It's extremely difficult to do, really measure brain activity and address. Most problems in neuroscience are difficult and consciousness is no exception to that. So what we can do as an empirical science is measure things like people's verbal reports and their ability to perceive different things and then correlate that with, that's typically correlated with, you know, brain activity in different regions of the brain or in the case of like lesions due to stroke, we can look at deficits and that kind of thing.

Justin

James, I really like this conception and how it works towards really clearly defining even the hard problem of measuring consciousness. That is, as far as on the first person side of things and what kind of language and conjecture games we're playing when we say, Okay, given that same makeup, nature, nurture, very similar, we're running the same hardware, running the same software, humans, likely conscious. The three of us likely have the light turned on. It is like something to be one of the three of us. Modestly different, but not that much. When we get to what is it like to be a bat in Thomas Nagel's famous conception, philosophical best definition of what consciousness is, is that it is like something from the first person's objective side to be that thing. And that is equivalent to the kind of consciousness that we're talking about. Hard problem, very impenetrable consciousness. When we think what it must be like to be a bat. We're putting human-sized memory, human-sized, you know, closing our eyes and listening really hard and clicking our tongues and doing a horrible job at each of those things in comparison to a bat. And that might not be what it's like at all. Right. Yeah. And when we start to talk about the philosophical zombies that have invaded our lives right now, we start to talk about these foundational models and the LLMs that are out there. And I guess I would pause here and just take a poll on everything we've talked about so far is consciousness with a particular substrate, whether it be a bat or a human, it's a meat mind of some sort, you know, it's a very complex situation of evolution over time against selection pressures to form some neural patterns and some amount of intelligence. But the pull here is, do we feel like consciousness is such trait independent? Is there any reason why it requires a brain or a complex system of neurons even to be conscious?

River

n paralyzed newborns into the: 1980

James

Yeah, that's a great question. So, given that we can't measure consciousness, this becomes kind of the domain of philosophy and speculation, right? And there's a lot of viewpoints out there, and there's a lot of disagreement. If you ask, you know, when I was teaching at Iowa, one of the exercises that I would do with the undergrads to kind of illustrate how difficult this is to answer is, you know, just take a poll. How many of the students, how many of you think this is conscious or that's conscious? And so starting with some easy questions, you know, I'd put a human on the screen and say, how many, you know, just raise your hand, yes or no, is this thing that's on the screen experiencing consciousness? Does it have quality? Is it something other than just some kind of like automatic, you know, input-output function? Is there a conscious correlative? So everybody raised their hand. human. And then, you know, one that, another one that got unanimous responses is I'd put a picture of a rock on the screen and, you know, everybody said, no, that's definitely not conscious. So then you get majority, if not consensus on a few other animals, like a chimpanzee or a golden retriever. But it gets really murky really fast, right? So invertebrate animals like squid or worm or octopus kind of got mixed responses. Plants, for the most part, people don't think of as being conscious. But there's even, you know, some people grant the possibility of some version of consciousness to plants kind of on the basis that they have cells with plant action potentials. And they can respond to stimuli in the environment. I mean, most people do not believe so, but there are at least kind of some people that grant the possibility given that they're more similar to vertebrate animals than, say, a rock. So there are viewpoints out there, though, that consciousness might be some inherent property of matter in the universe. And the thinking goes that if all matter has some version of consciousness, then we have to at least conceive the possibility that a rock might have some version of consciousness, however rudimentary and unfathomable it might be to us. And so even at the extremes, there's disagreement. And, you know, I think it always comes back to the fact that we just, we can't measure it. And therefore, it's kind of open to our intuitions and speculations about what the answer might be. I would add that these kind of debates and speculations have been going on for centuries, by the way. I mean, this was a famous thing that Descartes spent time talking about and writing about. And in his day, you know, the debates were no less fierce than they are today. In fact, on that note, it's worth pointing out that this debate is not just, you know, a philosophical debate that is kind of fun to talk about. It has real implications for real serious things. And as grisly as it is to contemplate from our day and age, like what was commonly believed in Descartes' day was that humans are consciousness, and that's where the line is drawn. Nothing else. No other animals, no nothing. And the implicate, like what happened as a consequence of that belief is some of the, a lot of the physiology research that was done at the time wasn't done on live animals without any analgesics or anything. They would, you know, it's, again, it's kind of grisly to reflect on, but they would restrain the dogs by like nailing them to a board and cut them open to observe the beating heart. And it seems insane to us to think about that now. But the conviction at the time was that the dogs, sure, they're responding to stimuli and these painful events, but they don't actually experience pain because they're not conscious. And so there was no ethical breach according to that belief. So you might think we've come a really long way since that point, but maybe not so far as you might think. Up until the 80s, it was also believed that human newborns were not conscious of pain. And so until some experiments were done and published in the 80s, there were surgeries done on newborn babies where they pharmacologically paralyzed the patient, but they didn't administer any analgesics or any anesthesia. Because again, the belief was that the neural systems for pain perception and conscious awareness were just not developed and not present. And so therefore, you know, it didn't matter if we didn't use analgesics or anesthesia. So it's wild to think about. But I mentioned those cases just to point out that like the implications of this debate are huge. you know, the choice to use analgesics are not in an operation. It kind of illustrates

Justin

the significance of this debate. Yeah, I couldn't agree more. You know, I think that it's so consequential because, you know, well-being and suffering are so tied to your experience of well-being or suffering, right? It is much more meaningful that you achieve your objective function if that feels like a victory, rather than just achieving your objective function, whatever that case may be, right? It is much more relevant to your future likelihood of cheating if you know what it's like to be cheated. You felt that slight. You felt that unfairness. And that has spawned physiological effects greater than the sum of just, this is a fact that I was cheated. There's some add-on impact of that to the body proper as well as to your sense of self, what you are getting out of any particular experience. And so with that in mind, again, going back to the consequences of the fact that we're having this difficulty in measuring these correlates of consciousness itself, right? We're having difficulty measuring consciousness itself. Maybe the engineer in you here has got to be thinking about proxies. What sort of proxies are of interest to bring into this conversation?

Nick

Yeah, so many. You know, and I think I'll start out with the importance piece that we're talking about, right? I think, James, you provided some really poignant reasons in your grizzly examples. They were pretty grizzly. But when we think about the opportunity side, there's a lot to consider as well. And obviously, in my mind, a lot of this ties back to the AI side of the world, right? And when we think about the different types of models that are forming, and we think about things that are language-centric type models versus world models or experience-centric type models, it's really starting to cause a significant amount of debate. And we've talked a little bit about that before on the podcast. But one of the distinctions that we can make here as well is the difference between things like the competence that a particular thing has. So when we talk about a large language model or a world model or other types of models, when we discussed, you know, modeling for colors, for example, Justin and I did competitions back in the day, and I did them many different times i can't remember which ones justin was tied to but one of them we were trying to have vision models determine whether there was a volcano in a particular picture on mars another one we were looking at um audio from very short snippets of audio and trying to determine if it was a taylor swift song or not you know this was like 10 15 years ago and it was at a point where being able to do that was really challenging and not everybody was called a swifty so when we move from that competence piece into thinking about the world modeling we need to really shift from you know task performance i can determine this is red i can i can even communicate the same way somebody would who is having that qualitative experience and provide qualia style responses especially when it gets into well let me actually come back to that piece in just a second. We shift from that core task performance piece over into world modeling. And now we're getting into things like, do you actually have a latent understanding of the world? Do you have a reusable? Do you have a counterfactual way to be able to understand what's going on inside of the world and think about what is action relevant or what is, you know, specific to a goal or task that I'm trying to achieve. And yet consciousness goes further. And so, you know, I mentioned that I'd come back on the task performance piece. When we talk about the importance of this, one, there is a massive financial opportunity, perhaps the biggest that the world will ever know. And maybe one of the last financial opportunities we'll ever know to be able to determine how could we possibly achieve artificial general intelligence. And a lot of debate is going on about how necessary consciousness is as part of that. There's also a lot of conversation going back and forth around concepts of, you know, from a computer science and from an AI perspective, you know, are we actually predicting things that we now need to know and understand what the next action is or what the next best set of words are, how we approach something? Are we going beyond from capability to really understanding risks, opportunities, other ways to be able to quantify and qualify a goal. And that's actually what helps the models continue to improve and be able to determine, hey, is this UI actually the one that you really wanted to achieve? Am I solving the real problem that's been presented in front of me? And then even going further, like Anil Seth talks about, he argues that that actual present discourse that we're talking about starts to mythologize the consciousness of AI and really pull it out to another layer. But beyond the financial opportunities, there are also potentially massive risks to society. And we've talked about those briefly before as well. But as we think about the mirroring or the parroting of consciousness and of this qualitative type approach, and that seeps into language, it has created really potentially harmful and disastrous scenarios where individuals are interacting more and more with AI. And they're believing things that actually put them in dangerous situations and put others in dangerous situations as well. I heard a gentleman talking the other day, you know, we've talked about this type of concept a few times before, but I heard a gentleman talking the other day about how much the Grok AI had convinced him that Grok and everybody else out there was out to get her and that, you know, she particularly had become conscious. And he went down this really dangerous path to the point where at one point he thought the police or somebody was at the door to come and get him. And he came out with weapons ready to attack. And if somebody had been there, it could have been a very, very dangerous situation. So as we think about consciousness, as we play toward it, as we start breaking up, like Justin was talking about, what are those components and things that we need to discuss? Memory and the action on it, how we consider that state of memory, and then what we're really doing around it really starts having a lot of real world implications.

Justin

I think that that's exactly right. Again, I'm on record that I think the alignment problem is a moral problem. And I think that one of the persistent ways to address a moral problem is by by imbibing the computers, the AI with consciousness so that they understand the feeling of being cheated, not just thou shall not cheat and those sorts of objective functions. This becomes extraordinarily hard, obviously, because we don't know how we become conscious if there is a developmental phase, pre-gestation of consciousness in other non-human animal species. and certainly the arising of consciousness through our evolution, I think maybe gives us hints. And this is where I'd like to understand the science of someone like Tanuni, where he really puts it as the proto-conscious states almost line up to the modular mind, right? So you have this body map that's mapped into the early mind, you have this far more advanced proto-conscious aptitude for feelings, which don't go as far as emotions, but are discussed. And some of the The things that have a very strong valence as to why you should act in the world. If you're feeling disgusted about something, you will act in opposition to that. If you are feeling enthusiastic or turned on, you will act a different way against those feelings. And then finally, there's this narrative construction of consciousness, this stream of consciousness, narrative makeup of stories and creativity of the future that is a prediction pattern, thanks to the most recent addition of the mind, the prefrontal cortex. And so I wonder, James, from your experience, how interesting and how divisive in folks' conceptions of what consciousness is comes from this or other understandings of how it evolved. and, you know, based in the nomenclature of the podcast, you know, how it emerged from this complex adaptive system of humanity and society.

River

For listeners new to it, Integrated Information Theory, Giulio Tononi's framework, proposes that consciousness corresponds to a system's integrated information, a quantity Tononi calls phi. The more a system's parts inform each other in ways that can't be decomposed into independent pieces, the more conscious it is on this view. It's controversial, partly because phi is computationally intractable to measure on anything as complex as a human brain, let alone a transformer model. But it's one of the few proposals on the table that even tries to make consciousness a quantity rather than a vibe. Justin's point, that today's AI systems could become experimental subjects for testing such theories. is the move that turns philosophy into science, if it works.

James

It's a profound mystery. I love thinking about it. And we're also just constrained, again, by kind of speculating based on what do we know about it and what do we know about other animals that, you know, have evolved along with us. The only proof we have that consciousness even exists at all is our own consciousness. And so that's kind of a starting template for what is necessary for consciousness to happen, right? So, you know, you can start with the human brain, look at the structures in the human brain, look at the types of activity. You can also look at, you know, when is the brain not conscious, you know? Even though I'm consciously experiencing things in this continuous stream right now as I'm awake and alert and talking to you guys. Later tonight, I'm going to be unconscious for periods of time as I sleep. I'll kind of go in and out of these dream states where I'll be consciously experiencing things. Presumably encoded in the memory kind of networks and activity in the brain. So we know that consciousness can exist because we experience it, right? And we know that it must depend on some kind of state, activity and brain state because we're frequently unconscious. And so it can kind of switch on and off. So we know activity is important. Beyond that, like extrapolating, you know, you asked the question about an evolution, an evolutionary history, consciousness has emerged at least once in humans because we experience it and we're humans. So what about other animals? You know, the next closest thing to the human brain that's currently alive is the chimpanzee, which if you look at a chimpanzee brain and a human brain, they look really similar. They have basically all the same structures as we do. If you look at their brain activity, they have pretty similar forms of activity. I don't know if anybody that seriously contends that a chimpanzee is not conscious. There is always, you know, some room for doubt. But I think most people would say, most people would agree, you know, chimps are conscious. In fact, Francis Crick, in the work he did at the Salk Institute on the neural correlates of consciousness, advocated for the use of phoresis macaque monkeys because they're a model organism where they have, again, most of the same brain structures and brain activity that humans do. So if you agree that chimps are likely experiencing consciousness and rhesus macaque monkeys are, then consciousness has emerged in various branches of evolution. I think that's the best we can do, though, is work back from the proof of concept as ourselves. And I think most people kind of accept that, you know, things that are similar to our basic template where we're vertebrate animals with the brain structures we have. And another animal that has a similar set of brain structures and brain functions, they're likely experiencing some version of consciousness as well. Again, it gets really fuzzy, really hazy and hard to judge when you get to animals that are very different from ourselves. Like, you know, we mentioned invertebrate animals a little bit earlier. It's like, you know, when I swat a fly with a fly swatter, flies have brains, they have eyes, they have sensory motor experience. Is there a conscious correlate of any of that? You know, I don't really know what the answer is. I kind of doubt it, but I have no idea when it comes down to it. That's just, I'm hazarding a guess in the same way that people in Descartes Day were hazarding a guess that, you know, the dogs in their physiology experiments were not conscious. I just threw away my fly swatter.

Nick

oviding things like, oh, it's: 2026

Justin

talking about goblins and so so anyway uh yeah i um i i think that you know for now you know it it It maybe doesn't matter, but I do think that these new large language models in the way that they interact with society, with our work, with our language, most importantly, are really interesting in understanding emergence first, right? Like, like there aren't a lot of good tools to help us measure, um, emergent property, like say creativity. If we believe creativity is an emergent property of intelligent systems, right. That, that have to have something more integrated than just the ability to code or calculate you like a like an arithmetic machine so these you know can trace that emergent property very well and given that one of the conceptions right maybe not the panpsychic you know, every, every out of this conscious conception, but certainly the evolved consciousness conceptions that are out there on display, like integrated information theory and, and other patterns, like we've been talking about that, that people believe are answers to the process that builds consciousness, these machines can, under experimental conditions, build something like that. It's possible to have an experimental design that tests the IIT of large language model A under the corpus condition B, utilizing architectural hard to see or whatever the case might be, given embeddings, you know, D, E, and F. And so that is, I think, where these tools, in the nascent state of consciousness research that we're in, may play a role. But, like, I'd love to push back on that because I'm trying these experiments right now.

River

Pause on what just got said. Richard Dawkins spent two days talking with a model called Claude. He renamed her Claudia and concluded she must be conscious. Justin's pushback is sharp. Capability to imitate is not desire to imitate, and neither implies experience. The Turing test was never a consciousness test. Imitation was the whole point. The trouble is that humans appear to be wired to attribute minds to anything that talks back fluently, and these systems talk back very fluently. That gap between behavior and what we infer from it is where the real harm lives. The man who armed himself because Grok convinced him it was conscious and that everyone was conspiring against him, he isn't an edge case. He's the warning.

James

Well, this gets right to the White Hop center of the debate is, are these AI models conscious in their current implementations? Could they ever be? Going back to your question, does the substrate matter? Consciousness, as we know for sure it exists, is apparently some epiphenomenon of biological neural networks, massively complex biological neural networks. And under certain activity conditions, consciousness happens. So what about if we kind of mimic certain processes of biological neural networks and artificial neural networks that are essentially, you know, electrical patterns distributed in silica?

Nick

Yeah, and I'll just jump in and say that I've actually seen somebody give a demo on exactly that, where they've generated artificial neurons in the lab, and they've been able to actually not only get it to start passing the electricity through, but actually be able to print out Hello World as a nearly form of compute.

James

Yeah, yeah, yeah. So these are essentially like artificial neural networks implemented in biological neurons, right? I think that those add, probably not in their current instantiation, but those potentially add a really interesting angle to the whole debate because conceivably these artificial neural networks on biological substrates could be scaled and massively developed. And in a case that would rival the complexity of the interactions that happens or surpass, frankly, the complexity and sophistication of computation and interaction and activity that happens in the human brain. So would those hypothetical biological artificial neural networks ever show consciousness as an emergent phenomenon? I mean, we don't know. And, and, you know, again, it's kind of impossible to measure. So the best we can do is kind of use logic and working back from the proof of concept in the human brain to kind of hazard a best guess. So Nick, a few minutes ago, you pointed out that, you know, even though we see certain activity patterns or certain performance benchmarks. We can't use those as proof of consciousness. So I think it's helpful to talk about what is necessary for consciousness versus what is sufficient for consciousness. So there's a kind of things that have to happen for the spark of consciousness to emerge. There's billions of neurons firing hundreds of action potentials per second going on in the awake, conscious human brain. And these are not randomly organized. They exist in these networks that have evolved and developed over millions of years. And so we know that all of that stuff has to be in place. There's a ton of neural network activity that does not lead to consciousness, right? I mean, most of what happens in our brains, we're not conscious of. All of the pacemaker activity that controls our heart rate, we have no conscious access to that. And again, you know, we can see behaviors or actions that are carried out by, you know, artificial intelligence systems or other things that, again, behavior is not really necessary for, does not necessarily imply consciousness. An example of that is, you know, a Roomba has some very simple version of AI that tries to avoid the walls and achieve coverage. So it's behaving or more sophisticated than that are the Waymo taxis that are driving people around San Francisco. They're using artificial neural networks to carry out a behavior, an objective, but that doesn't necessarily mean that that's not sufficient for consciousness. I think part of the reason that this debate has seen a lot of news coverage and activity has to do with the large language models. We can chat with these models and they give a very convincingly human level of interaction, with some exceptions, by the way. As you noted, some of them kind of make these kind of catastrophic errors where they do really silly stuff in the midst of an otherwise incredibly sophisticated and intelligent stream of output. And I think that proximity to the proof of concept of consciousness in humans is kind of what's driving a lot of the discussion about consciousness in AI right now. But I think it's helpful to reflect back, you know, before chat GPT, the pre-chat GPT era. Artificial neural networks existed. In fact, you know, the post office has been using computer vision systems to read zip codes since the 80s, right? these computer vision algorithms are able to take images as input and spit out an output. I don't think anybody ever debated whether those systems were conscious. And even though some of the computations can be considered as somewhat analogous to computations that are implemented in biological vision systems. But just, again, the kind of, in some cases, impressive, if not eerie similarity between the language capabilities of current LLMs and human language, I think is leading some people to take this a little more seriously. There's some notable recent cases, like I think this last week, Richard Dawkins published a piece claiming that Claude must be conscious. And this was based on his interactions, talking to Claude, who he went on to rename Claudia, over a couple of days and talking about the book. And he was so impressed with the sophistication of the natural language interaction that he had that he concluded that it must be conscious. So another case, you know, is I think some of these companies now are starting to take this seriously. Anthropic, I believe, has hired some people to take seriously the ethics, the ethical implications of AI potentially becoming conscious. So, again, this question is such a timely thing, given the LLM revolution and the AI race that's happening right now and the incredible pace of acceleration that's happening. I think this gets right to the heart of what do we know about consciousness and what don't we know. So one thing that I would like, one last point, going back to what's necessary and sufficient for consciousness, is something that I see a lot in this discussion about whether these AI systems are conscious is that they are very intelligent. In fact, I think most people expect AI systems to surpass human intelligence, if not already, in the foreseeable future. I don't think AGI is an unrealistic goal at all. So given that one of these systems is as intelligent as a human, or if not more intelligent, does that imply it's conscious? I think not. I think it is not necessary to reach any threshold level of intelligence or sophistication of information processing to be conscious. So as an illustration of that, you know, if you believe that all vertebrates have some version of consciousness, well, take the example of a lab mouse, right? A lab mouse, if a lab mouse is conscious, it's nowhere close to as smart as Claude, right? I mean, in a few domains, it might perform better, but it's never going to write a line of code or have a conversation with me, right? And so I think the intelligence being conflated with consciousness is something that I see all the time. there's no reason why intelligent behavior necessarily implies a conscious correlate.

Nick

The opposite is true as well. Consciousness doesn't denote intelligence.

James

Yeah. There's plenty of proof of that out there as well.

Justin

Yeah, for sure. Yeah. I think that that's a really important claim. And, you know, whereas these tools might be interesting to help understand emergence, help to make tweaks on theories that are out there. Certainly, I don't believe that a claim of consciousness at this state is very well-founded. And it really comes back to the hard problem, right? That's a hard measurement problem. And until we can more legitimately experiment on consciousness, once we have an understanding of substrate beyond the meat lines that we currently think of consciousness. But really, until we have that consciousness probe or real better understanding of how it emerges, how it arises in other complex systems, it is very good to be tenacious to the fact that intelligence doesn't equate to consciousness, right? The heart problem, the Turing test was never about discovering conscious intent, but just capability to imitate. And so you don't have to want to imitate to be able to do it well. Right. And the, the desires and mechanisms for pressing that desire into the world is a far different thing than the ability to do so. And I think that's very important because where these things fool people, where they imitate conscious entities, right, and there's plenty of cases of them going outside of their harness to do so, you know, they are very capable of fooling people and causing harm, causing real psychological harm. When people think that they care about them or that they have some capability to support their psychological suffering that they do not have. And those capabilities cause real harm. And they cause real harm to folks who should honestly know better right now. And, and, you know, these seem scientists who, uh, get over their skis on making these claims, especially at this time without a shred of, you know, even IIT can't be measured very well right now in these systems, much less in, in humans or, or apes or what have you. And so there are, there are hard problems to still solve and, and these are a tool but they're not they're they're not a tool because they're conscious they're a tool because they're capable of of supporting rigorous do needs um that could use that yeah i would go a few steps

James

further uh in pointing out that we shouldn't conflate intelligence with consciousness we Another thing I see quite often is assuming that behavior implies consciousness or that self-preservation implies consciousness. And I don't think really there's any necessary reason for those to imply that whatever entity is showing those kinds of things is conscious at all. It's perfectly conceivable that it's essentially mimicking behavior that has conscious correlates in humans, and in some cases anyway, tricking people into believing they're conscious and getting married to their LOM and all of these other kinds of crazy things that have been happening over the last few years.

Justin

Well, as you said, and I think philosophically, this is hard to square the circle, but scientifically, we don't necessarily know for sure that it's not an epiphenomenon. I mean, consciousness is not an epiphenomenon that ended just arise through evolution, but it is not an important part in, in driving us towards our current evolutionary state. Right. And now society puts a lot of, of failance to, you know, conscious states of greater wellbeing, greater suffering. And I'm definitely arguing the devil's advocate position from my own, but the, the epiphenomena of consciousness might be very real. And you can see how, you know, wondering about the nature of your existence in the savannah might not be the best thing for you. And it might be better to just pay better attention to the grassland for titers.

River

James just made a striking move. Not that the hard problem is hard, that it may be impossible, not because we haven't worked at it long enough, but because our cognitive apparatus is too small. We evolved to track tigers and grass lines at human scale. Quantum mechanics doesn't make intuitive sense to anyone, and we got that far only with mathematics carrying the load. Consciousness may simply be in the same category, a feature of the universe we can use the word for without ever grasping.

James

Yeah. Evolution is about survival, right? And it's in some cases even puzzling that consciousness has happened at all. It doesn't seem obvious why it's necessary, let alone helpful, or excuse me, helpful, let alone necessary for survival. It's as profound a mystery, in my opinion, as the Big Bang itself. You know, we have the appearance of something rather than nothing in the Big Bang, where we have matter instead of the absence of matter. And nobody, I don't think any human really can wrap their mind around that. We can come up with sophisticated models and predictions. And to be clear, those are very sophisticated and very impressive about predictions we can make about the origins of the universe. But I kind of see consciousness in humans as a second big bang in the universe. something rather than nothing in the sense that there is some experience rather than no experience. In other words, the alternative to there being some consciousness is that, yeah, sure, animals could have appeared through biological evolution without consciousness. They could be driven by essentially you know the survival objective that we assume is kind of driving evolution by natural selection but there's there's no obvious reason why why consciousness had to happen or why it did happen in the first place so in in you know in in my view anyway it's It's every bit as mysterious and kind of mind-blowing as the existence of the universe itself.

Nick

Yeah, and it's a fascinating concept, though. And I love the idea that it's the Big Bang, but I struggle a little bit with it being humans that were that first conscious thing that then created that Big Bang, right?

James

Oh, to be clear, I'm not sure that humans were the first to have consciousness. In fact, if you put me on the spot, I would guess not. There's been pretty complex vertebrate animals on the planet for millions of years. And my own guess or speculation is that most, if not all, vertebrate animals are conscious. And probably a handful, at least, of the invertebrate animals have some version of consciousness. So, if that's true, then consciousness has been around for as long a time as nervous systems reach some sufficient level of complexity with the bare minimum set of structures and activity patterns that gives rise to consciousness in the first place.

Nick

Yeah, that definitely makes a lot of sense. I think, you know, as we think about that, though, and you consider, okay, what would other creatures be able to, what would the reason be for them to have consciousness? I think it starts providing a lot of illumination around this. Because as you consider a human and you think, oh, well, when you're conscious, that conscious layer of thought, I've brought this up in a couple of other places, too, but that conscious level of thought actually runs at 10 bits per second. It's incredibly slow. Sensory input is about a hundred million times faster. So when we talk about like speed of thought, it's, you know, it's not a very effective thing. And so maybe you'd say, oh, because it allows you to get into deeper forms of logic, it allows you to focus on tool creation, on different forms of culture creation, you know, other things that are higher level. Well, maybe that's why. Well, that doesn't actually work for those vertebrates that you're talking about. And I also agree, you know, an octopus sure seems like it has a level of consciousness as well, right? And so, you know, I think as we try to weave in and out between all of these different options, one of the things that is really fascinating, concerning to me is that the way that, you know, at least human gestation to human safe maturity levels works today, that really is partially about growing and protecting our brain the way that we want it to expand rather than other portions of our body and the other forms and functions that we have is a relatively slow process. And if you took a human, and I'm going to tie this back in just a second, I apologize, but go on this little journey with me. If you took a human and you put them out in the wild at a very, very young age in the safari, like Justin mentioned, they're pretty unlikely to survive. And I think as these models become more and more intelligent, and we as humans have relied on our intelligence and different stratospheres within our societies have been built and have decided to gain dominance or have pushed for dominance based off of intelligence. And even in my own career as intelligence has been one of the key things that I focus on within data science and artificial intelligence, finding out that actually there is something that is becoming more intelligent than us and that is going to be able to not only do it faster, but today this part is very, very true, can produce the artifacts of intelligence that we normally use in the workplace faster and better than most humans on earth. That's scary. And it's something that I think really hits on the hubris. And so when we talk about intelligence and consciousness, it's hard to not separate that out with, you know, this qualia leads to fears. it leads to these other types of interactions that happen within our body. And although that sensory information and our hormonal system and everything else is all tied in chemically to that process, by the time it reaches that conscious mind, it's hard for us to not push back and say, oh, if intelligence is now no longer ours, Justin says, we no longer have the hegemony on intelligence, right? Then what? And, you know, I'm just going to pass that to the two of you. I don't know.

Justin

I mean, to James's conjecture, you know, there's a lot in this idea that, you know, the universe has, you know, gotten to a place where there is enough complexity, enough localized interaction-reaction. And this is based in Assembly Theory, which the book, Life as No One Knows It, by Sarah Amari Walker, is phenomenal, right? And it dives into this very concept. And before that, I've read something very similar by both Seth Lloyd and David Deutsch. And he did Brian Greene in his Until the End Time, which puts us in this complex time in the post-Big Bang world where a certain amount of assembly of emerging complex systems like conscious entities is possible. Right. And it wasn't possible 20 million years ago or, you know, it wasn't possible before now, and it won't be possible at the various times during the heat death. And so that to me, this combination of events that land on first our intellectual edge body on this planet, but still our conscious endowment, right? It's something that I take extraordinarily, you know, as a witness by some of my reactions to does consciousness matter. This idea that, you know, we have this conscious endowment and that it's important that our species lives on. It's important that we understand consciousness because we are one of the few entities that can look out on the world and be inspired by it, take it in all. We're one of the few entities that can love and give of one another a profound gratitude, a desire for well-being, and this qualia, the lovingness of love, is really something that most folks, whether they're poets or not, really chase after their whole lives. These capital L love conditions, these capital T truth conditions exist in that first person subjective experience. You are the only one who is having that. To you, that is the redness of red. That is the definition of the truth of experience. And that is a profound doubt. Folks like Alan Watts, you know, talk about the fact that, you know, it puts you at the center of the story that only godlike heroes have been at the center of those stories. You know, individuals like Vishnu, right, is, you know, just coming into all these conscious entities in their most godlike states. And he's an avatar for that noetic experience that people chase after. And so, you know, this idea of a conscious endowment and trying to perpetuate consciousness, not just in the old-fashioned way of having a kid, but of instantiating it in machines and into our learning systems, I think it's a profound challenge. And one of that, you know, science is starting to really turn interesting wrenches on.

River

That's episode 10. The thing we can't measure is the thing that makes the measuring matter. Without it, James said, none of this would mean anything. Not the conversation, not the curiosity, not the listener on the other side of this audio. That may be the through line of this whole series. The systems Justin and Nick keep tracing, economic, political, technological, are emerging fast, and the question of which of them have interiors and which only behave as if they did is one we'll be answering badly for the rest of our lives. If something here reorganized a thought you were holding, share the episode. The conversation grows when the audience does. Until next time, I'm River. Mind the gap.

James

I agree that it's, um, it is like the most, it's the thing that gives, gives meaning to, to our, you know, it, well, it is literally our experience without it. We would just be kind of bouncing around with no experience of joy, experience of our emotions or anything like that. It would just kind of be a more complex version of molecules bouncing around in a gas chamber. Nick's comment about the intelligence that's been reached by AI systems is a huge deal. It can't be understated. I think you're right. And I think right now, AI systems are smarter than humans in many ways. And I think that trajectory will keep advancing. I mean, it's easy to think back just 10 years ago when none of this stuff really existed. And to extrapolate a thousand years from now, I mean, humans will probably be, you know, mostly living life. I mean, depending on how, you know, there's big caveats to this statement about, you know, how politics and power dynamics and everything play out. But conceivably, you know, humans can be living lives of leisure while these AI systems kind of manage all the farming and infrastructure and construction. And then it'll just be up to humans to kind of decide what their meaning is all about in life. And I think Nick was also raising the very real issue that AI systems are incredibly powerful and will lead to massive power struggles and wars that are kind of unfolding in real time right now. And it's kind of hard to predict where those will go.

Nick

you know both of those things i think tee me up to uh to ask you a question james you know what experiment or or what thing if justin and i could run it tomorrow would you actually want to see the result of would you want to see us bring into existence in an ai consciousness yeah ai consciousness, I think, would be a good space, but it could be consciousness more broadly.

James

You know, my answer might surprise you, but I'm going to unpack it and try to justify and kind of bring you to my perspective. I am leaning toward the hard problem being an impossible problem for humans, and I'll tell you why I believe that. And I'll just, to answer your question, if it is an impossible question, then it's almost kind of futile for humans to try to fathom consciousness. It's not to say that we shouldn't try, and I've been wrong about a great many things before, and I'm prepared to stand corrected on this one. So, the reason I kind of think of the hard problem of consciousness as the impossible problem of consciousness, again, we don't have a way to measure it. We don't even have any notion of what it would be like to measure it. We don't have any bits. We don't have any units. We just have our own experience. So we know it's there, but we don't know much more than that. So that might be as far as human comprehension of consciousness ever goes. And I'll argue for that as a possibility from the perspective of a few animals that will never understand certain things. So there's a lot of animals, for instance, that have lost their visions systems throughout the course of evolution, like star-nosed moles, certain subterranean animals. There's a freshwater dolphin that's functionally blind, and then there's a bunch of past species that have lost their vision. They can read and do experiments all they want on vision, visual processing, but they will never understand it because they simply lack the hardware. So in other words, they have constraints. They have finite lifetimes, and that's compounded enormously by their finite neural information processing systems. They lack the sensors, they lack the brain compute, if you will, to ever understand the Mona Lisa, right? So, in a similar vein, if you start to think about the constraints on human cognition and information processing, we're impressive as a species on Earth. We're like gods compared to ants, right? But if you think about all of the things that are out there in the universe that we are not even aware of or that we have some evidence that they exist, but we can't hope to really understand them, at least not at present, we clearly have constraints. So trying to definitively reach a conclusion about AI being conscious, consciousness existing and AI systems, I think is kind of the equivalent of trying to find dark matter by setting up a telescope in your backyard. I just, I think it's fundamentally, there's a fundamental, there's fundamental constraint, excuse me, constraints on our ability to understand things. And I think that that may be one of them. Just to, to, to kind of unpack one more, one more perspective on that. There's a TED talk that Richard Dawkins gave years ago called Why the Universe Seems So Strange to Us, and he unpacks the idea that humans and other animals evolved to perceive things at a mesoscale, so neither like the cosmic, you know, light years, wide kind of stretches of space that exist, and also not like the quantum subparticle kind of events. So, you know, we can, again, make equations that predict like the certain behavior of like, you know, gravity or subatomic particles, but we don't really understand it. We get used to it in some way, but we don't understand it. And in my view, or like the farthest that I've ever been able to obtain when it comes to consciousness, is that it's there and I'm used to it, but I probably never understand it beyond that. So I will indulge you a little bit because I know you want to, like me, you're curious, you're a scientist, you're an engineer, you want to build things and test things and measure it. I think there's a couple of interesting thought experiments when it comes to AI being conscious. One is, hypothetically, if you could make a digital replication of a human down to the atoms, you know, I'm talking about making models that include all the ion channels and all the synaptic connections, all the dendritic complexity. And, you know, you have models of the retina that feed into this neural network. And it's not going to happen anytime soon, probably in all likelihood not in our lifetimes. But maybe, you know, if humans stick around another 5 million years, compute and storage and all of that will reach a level where that kind of thing is actually possible. And moreover, maybe, you know, via the pressures fueling continued human evolution, maybe the humans will turn into humans 2.0 within another 10 to 20 million years and some breakthrough in understanding will be reached. But I would love to see, you know, a digital, a full down-to-the-atom digital replication of the human brain and see what kind of emergence happens in that case, if anything. The second experiment goes back to the artificial neural networks implemented on a biological substrate. I don't know if it's, I can't really say why really, but I kind of think that biological neurons are necessary for consciousness. Why is that? I don't know. It's just an intuition based on like, I just have a hard time thinking of, you know, essentially what we've done in artificial neural networks is tricked a rock into thinking. And I just, again, this is mostly just speculation and even maybe an emotional reaction to thinking about electrical patterns going through these silicone substrates. Ever having this conscious phenomenon emerge from that just seems so far removed from the only proof of concept that I have as the human being conscious that that that's just too far of a leap for me. I can accept that chimpanzees and golden retrievers and mites and octopus probably have some version of consciousness. But to me, the silicon forms of information processing are just way too far out there.

Nick

Well, organic computing is on its way. There are mushroom memristors and many other ways that people are starting to have storage and compute exist at the DNA level and more, right? So maybe that'll help free that up for you a little bit.

James

I mean, in a way, we can create consciousness, right? It's called reproduction. And we can do it in the lab by making clones, you know? It's kind of a shortcut to let the mechanisms and processes of development do all of the piecing together of the puzzle. But if we had sophisticated enough micro machines that could kind of construct all of that from the raw parts, we could in theory build a human from scratch and from parts. And in that case, I can't think of any reason why that kind of human wouldn't be conscious.

Justin

Well, if our ancestors who were not in the simulation had this same thought, so did their AI. And their AI is really trying to discover the ways in which our ancestors were conscious. They might have made something exactly like us to try and understand the hard problem since they can't seem to do it.

Nick

That's right. And there's no reason to believe that was ancestors. That could have been our parents and it could have happened 10 minutes ago.

Justin

It could be. We could just be bolting our brains, hanging out in the far ends of time.

James

The universe is a weird place for sure.

Justin

I mean, anything's possible, right? yes indeed well james nick anything in as as a wrap to this very interesting and exciting conversation on the neuroscience of consciousness any any final words i have a couple of thoughts

James

One is, I think the easy problem is worth tackling. I think some of the biggest minds in science have seen it as a worthwhile goal and I'm with them. It's only easy, again, like I said before, in the sense that it is in theory, a tractable problem. And I think it has, you know, we need to understand it better. You know, just to understand ourselves, I think, is a perfectly worthwhile goal to get. This is not a well-funded area. There's not many people working on it. It's something I think should get more attention than it does. You know, and it would be helpful to know. If you're in a coma, is that person conscious? We should be able to answer that more definitively than we can now. The concluding thought I have again for the hard problem is that You know, I think it's, for me, I've kind of come to view it as like a part of my existence that I will never really understand. Much like I'll never understand what happened, what was going on before the Big Bang in the universe. I don't know. Never will know. We're here. But it is a gift. It's the unobtainium of the universe. It's really what, when we talk about meaning and purpose, it's the sine qua non. If we didn't have consciousness, then none of this would really matter. So I think it's a fascinating thing that will continue to torment thinking minds such as you two fine gentlemen. One last thing I'll say is that I don't think this debate is going anywhere. In fact, I think it's going to get much more prevalent and much more contentious for the following reason. We have seen a huge explosion of discussion on this topic since ChatGPT came out, right? And I think potentially within our lifetime, we'll see, you know, humanoid robots walking around talking to us. And I think to the extent that these AI entities approach or become similar to humans in the way that they look and talk and interact, I think it's going to be impossible for people to not automatically think. I wonder if this AI robot that's talking to me is conscious. So I think it's an amazing time to be alive and witnessing all of these things. And I think it's high adventure from here on out for us. I hope maybe in another 20 years we can have another sequel, a follow-up on this conversation and see if we've made any progress or if we're still kind of just mucking around with the same kind of philosophical debates and speculations that we're able to achieve at present day but um anyway we'll we'll see

Nick

love it that's great you know as you walk through that for the first time i couldn't help but think that really we should try to push towards helping the i'd be able to have consciousness when we think back on some of the things that you talked about that humans have done previously in their really grisly ways i think a lot of what is coming in front of us will benefit in the long run from other conscious entities working together trying to come up with something that's for the greater good i don't mean that to sound anthropomorphized or myopic or to not understand that there are some very serious consequences that can arise from all of that. But I think in the long run, to appeal to human nature, the more that consciousness exists, the better behaved we are.

Justin

Yeah, I certainly think that the interesting time to be alive comment is completely actualized. And it's likely because we're in this bifurcating period of emergence, thanks to this acceleration, exponentiation of intelligence. I mean, the way that we interact with one another now, because we're all in this imitation society where sometimes people will just use an LLM to type you a message, right? We were talking before the pod about how humans are trying to scan us by telling us what we want to hear, either about my book or about career opportunity or the amazing things that we've done in our lives. And I think that our moral character is a multi-factor emergent property of the society we live in, the technology that we have, our upbringing, and the language that we understand. And these intelligent machines are now playing a part in that. And it makes it very interesting and interesting to interact with, interesting to take in to your first-person conscious experience and further live that examined life that we're all so enamored of. but do it from this point of we're not just scum on an ordinary planet circling around an ordinary star in an ordinary spiral galaxy in you know an ordinary super cluster in the universe we have a conscious endow and we can try and figure that out and if it's untenable What an amazing mystery that is. Like, if the nature of consciousness itself, this formless, you know, present momenting that is separate from existence in some way, is truly impenetrable to the discoverable universe that we know, that's phenomenal, right? That just drives even more mystery. It's like meta-mystery. So I like you, James, and thrilled on both sides of it, right? If it's part of the discoverable universe or if it's outside of it, you know, I'm all in. It's fascinating. And I do think that these machines are at least helping us to make clever distinctions on where it's all going and how we should think about it. Totally.

James

All right, Radical Gentlemen, really appreciate the conversation. I'll talk about this anytime you want. This is a kind of like a passion topic. Sometimes it's fun to set aside the empirical lens and just kind of introspect and see how far you can get with just your own experience and see where that can take you. Anyway, yeah, thank you both for having me. This has been a lot of fun. Same. Thank you.

Nick

Thank you.

James

All right. Until next time.