Conversation #1 with — Thoughtforms Life Podcast

This transcript is automatically generated; we strive for accuracy, but errors in wording or speaker identification may occur. Please verify key details when needed.

Main Episode

[00:00] David King: One of the bits of the conversation that I'm looking forward to the most, and maybe getting ahead of myself by leaping into it immediately, is the idea of sort of autonomous, agentic, nonconventionally human formal structures sort of flowing into atypical configurations of human matter, and what that looks like and feels like from the inside and how it appears from the outside. Particularly, as someone who's training in psychiatry and working on the inpatient units and seeing people who are reporting experiences that are of precisely that nature. And because of the inability to kind of hold that worldview in a way that would benefit the patient within the sort of frame of understanding that we have within psychiatry, and it wouldn't necessarily serve any clinical benefit to adopt that view ourselves, we tend not to. We draw back to our more sort of conservative expectations that these experiences derive more from the transient malformations of a psyche with too much dopamine in the wrong circuit rather than the idea of there being an actual influx of agents into a space. And it's still such a, I don't know what the best term is, heretical area of discussion psychiatry that I feel myself a little nervous to talk about it, with the knowledge that the conversation may be viewed by a large number of people. But I think it's an important conversation, and I think it's something that we should take seriously and at least sort of work through from an intellectual perspective and see where it sort of holds water and where it leaks. And, yeah, I think our patients deserve the conversation, at least.

[02:41] Michael Levin: No. That's very interesting. I mean, I think from the perspective of putting things on rigorous footings, I would say that the framework from which these ideas seem heretical and ridiculous and all of that hang together with a bunch of assumptions that are demonstrably false, given modern science. Well, actually, given science of a long time ago, but certainly of modern discoveries. And I think we owe it to everybody to shed a light on that and update things because our conventional expectations and assumptions of what makes sense and what's silly and so on are really bad. We're not calibrated well for what we now know is the facts on how biological systems function. So, yeah, I think, you know, I'm sure that this is—I mean, all of the stuff that I talk about is completely heretical in my circles too. You know, we've all learned in biology class that the right way to handle biological systems is as a set of chemical rules where nothing knows anything and nothing has any goals. And that was helpful a couple of hundred years ago getting science off the ground when you could either be dumb like rocks or smart like angels and humans. So then fine. Let's go this way and see if we can see how far we get with everything being dumb and nothing knows anything. But we have a science of cybernetics now, and we know that systems that don't know anything and don't have any goals are just the lowest rung of a well developed hierarchy of systems. And the fact that we have to stick to that when we talk about cells and tissues and whatever is an assumption. It's an axiom that people sort of bolt on as if it were necessary, but it actually is not only necessary, but I think demonstrably false. And so, you know, yeah. And so people immediately, you know, have this reaction to the things I say as a fullback. I mean, that's crazy. That can't be. Like, well, let's just dig in and see what makes you say that. What is that based on, actually? And it turns out that, alright, it's an assumption we took on, you know, prior to having a science of these things. So I think we're gonna find the same thing here. And like you said, I think, yes, it sounds wild at first, but we have to start off with the right, we have to recalibrate our assumptions, I think.

[05:06] Michael Levin: Yeah. Right. You go ahead, Dave.

[05:12] David King: You're about to go, Tom.

[05:14] Thomas Pollak: Well, thank you. I was just gonna say, Mike, was reminded a little bit when we started working on the AI-associated delusions paper. One of the reasons that we got very excited was this notion that you had that maybe when you have an agent in your ear, there's a certain amount of if you have this always-on agent in your ear, and because initially, we're thinking of people with psychotic disorders who sort of, in a sense, coexist with other agents who are sharing some sort of mental real estate with other agents, no matter what you how real you think they are, their experiences of that. And you had this wonderful idea that we kind of had to expunge from the final paper a little bit, although it's still there in minimal form, that in this future when we have this agent sort of in our area, whether it's Scarlett Johansson or whatever it is, there's gonna be a sort of battle for real estate, and one possibility actually is that there just might not be enough room for the other agents, as it were. And there's kind of a very paradoxical implication that maybe this could be a good thing. And, of course, we have no way of testing that yet quite. But it got me thinking, I think, something David said in one of his emails that

[06:44] Michael Levin: I

[06:45] Thomas Pollak: I don't think it's quite that it's a zero-sum game, but this idea of a limited amount of real estate feels like it could be quite generative because I think we've all said in our own ways that disease in general and certainly psychiatric disease is pretty much universally characterized in a meaningful way by a loss of agency for the person for whom it happens. And if you are, is there a way of thinking about it where the total amount of agency in the organism is roughly the same? It's just that the locus of agency is sort of redistributed. And so in a way, there are these perhaps low is not the right word, but these other sort of aggregations of agential patterns that temporarily have more agency than they would normally, and the result is that you as Tom or as the patient have a reduction, but the patterns have more in there. And consequently, part of the therapy is actually to help redistribute the agency within the system or kind of enlarge the system or something. But yeah.

[08:02] Michael Levin: I mean, there's also, there's even a wilder version of which I think I might have mentioned to you, Thomas, which is that if it is true that, I mean, one way to start thinking about this is, like, we know, for example, in mammals, at least, when you shut off visual inputs, then other algorithms will move into the visual area, right? Hearing and things like that will take over new real estate. So for me, I often think about, well, in embryogenesis, I could give you a third hemisphere. If you're computational processes are good at taking over areas that aren't doing anything, we could have more real estate if we want to. So in general, this idea that these kinds of processes like to have substrates to run on, so to speak, and I think there are other interesting ways to support that, then you could even do something even crazier, which is provide a host hardware, so to speak, for some of the things you would like to get rid of. So whether they be obsessions or repetitive thoughts, or bigger, even bigger things possibly as much as fragments, personality fragments, whatever. And then this is like, I'm talking beyond my psychological expertise, of course. So actually provide real estate for them that can be disengaged. In other words, now I'm saying, okay. I have this set of goals that live in me, these OCD things. I'm not gonna try to kill you. I'm not going to try to get rid of you. I'm going to give you a host. It isn't gonna be quite me, but maybe we all live together or something. You live in this thing. And, you know, I'm gonna go to work and have a nice day. You, if you would like to, keep washing your hands or better yet, you sort of sublimate that into cleaning the kitchen, like, nice. Go ahead. And you can do that, and you will have real estate. Like, you will have the real estate. You know, go for it. And create some sort of maybe it's a BCI, some sort of brain-computer interface. Maybe it's a full-on, you know, kind of prosthetic that's not attached and can be detached and set aside. And that may be another way to deal with some of these things as instead of fighting them and trying to stamp them out, actually give them room to sort of play out whatever it is that they wanna play out. And you can imagine all sorts of technologies for that kind of thing. You know?

[10:31] Michael Levin: I I think

[10:31] David King: This is so interesting. I the discussion of internal or intrahuman polyagency seems more palatable than evoking some kind of third-space formal entity that maybe enters the picture. So it makes sense to start where we have. If we take seriously that the human organism is this incredible nested hierarchy of minds, each with their own sort of individual cognitive lacunae, and then those that emerge from local collaborations and so forth. All of these different goals and intentions and set points and a lot of competition as well as collaboration within this. And you're kind of describing a very rich but highly political ecosystem competition. And one of the things I'm very curious about is what happens when you have a sort of rogue agent or rogue mindset that's able to travel vertically or commandeer apparatus that exceeds its own ordinary playpen for its own ends. So thinking, for instance, about substance use disorders where a relatively minor bit of the cortex completely takes over the behavior of the organism. So maybe it's a good point in the discussion to ask what communication within the internal hierarchy looks like. How does that happen? And how do these negotiations actually take place at a bioelectric level in a way that subsume or consume or associate or disassociate?

[12:46] Michael Levin: Yeah.

[12:47] David King: So

[12:48] Michael Levin: So, okay. Right. So there’s a couple of things. First of all, the only kind of two things that I think we know for sure are that questions about agency of anything is an empirical question. So I don’t think we, you know, when people say, well, come on, that doesn’t do the trick. You have to actually do experiments and use the tools of behavioral science to analyze. And we’ve analyzed all sorts of wild stuff, starting with gene regulatory networks, random electrical circuits, predator-prey, lots of other kinds of ecosystems. All of these things have different degrees of cognitive competencies, and you don’t know about them until you actually empirically test them. You actually do the experiments right out of the sort of behaviors-to-handbook. You just do the experiments, and you say, oh, wow. Look at that. This thing can do Pavlovian conditioning. You know? Who knew? So we have to, so these are all experiments. We can’t just kind of assume anything. And then the other thing is that we are fundamentally collectives. So there’s no getting away from that. When people say, well, come on, you know, me, me, I’m this indivisible thing. Well, everything is made of parts, and there are lots of questions to be asked about which of those parts get access to the language and the wallet. So that, you know, and this comes up a lot when somebody says, I’ve anesthetized the patient. Well, you anesthetized the one that complains about it afterwards and sues you. He’s anesthetized, but I don’t know what else is in there that doesn’t have access to the verbal part. And just in general, I’m very skeptical of, like, you know, when people say unconscious processes. How do you know? Well, I asked the person, and he, right. That’s great. It’s all calibrated against what the left hemisphere has to say. But what do you know? So with all of these things, we have to do experiments. We can’t assume anything. And I’ll just tell you what we know about the scaling of the self. There are a couple. I’m sure there are others, but we’ve identified two kinds of what I call cognitive glue. And that’s just basically mechanisms and policies that align competent subunits in a way that gives rise to some kind of a higher-level agent that knows things the parts don’t know. And what that thing is constantly doing is bending the option space for its parts so that they can keep doing what they know how to do, but it all serves a higher purpose that they don’t know anything about. And so bending that option space by hacking them, by behavior-shaping them, by rewarding them, is what it takes to be a higher-level self. You have to be, first of all, in charge of your own parts. And the way that, so here are the two that we know about. There are probably others, but the two that we know about. The first one is memory anonymization. So what happens is, just imagine you have two cells. I poke this cell. It generates some kind of calcium flux as a memory of it. It releases some sort of signaling molecule. That signaling molecule comes over here. This cell knows perfectly well that that didn’t come from me. It came from outside. So I know that’s not my thought. That’s your thought. And I can ignore it. I can do something with it. I can take action or not. But we’re very clear. What are your thoughts and what are my thoughts? What happens in the body is that most cells, or many cells, are connected by these gap junctions, these electrical synapses. And when that happens, small physiological signals, so electrical current, but also calcium and ATP and things like that, can go from one cell to the other. So now this is a very different business because if I now get poked and my calcium engram ends up in this cell, I no longer know, was that your thought or was that my thought? Because we encode them the same way. And so in some sense, it’s a false memory, but in another sense, it’s a very true memory of what happened to us. And so there used to be me and you, now there’s just we. And, yes, we got poked. And in fact, you know, we got poked on the left side of the body, let’s say. Just like when you train a rat to press a lever and get a reward, there’s no single cell that has both experiences. Right? But the rat knows, and the rat knows because it has this system that anonymizes who the, you know, the information is shared, basically. So when you have this thing where you have this mind meld that you can’t quite tell which memories are mine, which are yours, then now there’s a we.

[17:13] Michael Levin: That's the first thing that we see. And the second thing we see is stress sharing. So stress sharing is basically just being leaky with your stress molecules. So if you have some kind of, and we can measure those now. And when you have some amount of stress because you're off of whatever set point you want it to be at, if you just let that stress leak, it'll stress out the other cells around you. And, fundamentally, your problems become their problems. Not because they're altruistic, not because anybody wants to help, but just because you're stressing them out. They're using the exact same stress marker as you. So everybody's temperature sort of goes up. Everything gets a little more plastic. Everything sort of moves around. And then once everybody's happy, the whole thing can settle down. So it incentivizes alignment and collaboration just with a really simplistic mechanism of just, you know, leave. This is a stress leakage. And we've modeled all this and shown what the implications are for some of the questions we're interested in. So what I think is happening is that because of stress sharing, memory sharing, and probably other things, there are larger scale subunits forming in the body at all times. Some of them are relatively stable. Some of them can grow and shrink. The basic process of growth is embryonic development. It's when a, you know, you look and you say, well, there's one embryo. What is there one of? There's a, you know, a million cells. What's there one of? Well, what there's one of is a kind of shared delusion. All the cells are bound together, and they've all decided that we are this magnificent thing. We're gonna be a baby giraffe or a human or whatever. So we're gonna make that journey from being a single cell in more of a space, in the anatomical space. We need to go over here. And they all agree on that. And because they're all electrically coupled, if you cut them in half, then they'll do it separately. And so you see the scaling of the cognitive light cone where tiny little goals of individual cells, pH, metabolic, hunger level, those kinds of things, become suddenly scaled into very large set points that huge networks can remember. The shape of a limb or the number of eyes. You know? The single cells can't remember that, but the collectives can. So during development, you get this growth of the cognitive light cone. During cancer, it's a dissociative identity kind of disorder because individual cells disconnect. As soon as you disconnect, you can no longer access this big memory of what we were building a kidney. Now it's like, well, I'm an amoeba, and I do what amoebas do, and the rest of the body is just the external environment as far as I'm concerned. So it can shrink, and it's very plastic. We've shown you can scale it up. You can scale it down. You can force cancer cells to reconnect and be normal just by virtue of electrically connecting to their neighbors. So this whole thing is plastic, then you have interesting dynamics. Like, for example, we have this work in the slime mold. Right? So the slime mold is sitting here. You put a piece of oat. The slime mold starts going for the oat. Very obvious. Well, I'm gonna get the oat. You take a razor blade, and you cut off the front edge of it. And now it has a decision to make. Well, I can rejoin the collective, and then we will go get this oat. Or why do I wanna share this thing with this giant mass spec here? I can keep going and get all the nutrients for myself. But that wasn't available to you any more than the metastasis decision was available to you while you were connected. When you're connected, you're in mind mode. You can't even have those thoughts. But as soon as you're separated, oh, now the payoff matrix, the, you know, the cooperative major payoff matrix changes. And now decisions you make about your and we actually did work modeling a spatially distributed prisoner's dilemma where you could defect, cooperate, merge, and split. So that was cool because whenever you do one of those other two operations, it completely changes the payoff table after that. Right? Because the number of players is not constant. So that's what I think is happening here. The material is extremely plastic, which starts in embryogenesis. The boundaries can sort of grow and shrink, and these boundaries exist in many spaces. So maybe these boundaries can do that in, you know, sort of conventional human cognitive space. So that's maybe how you get dissociative identity disorder that you guys, I'm sure, deal with. But we see it on an anatomical scale all the time. So I think a lot of this is about learning to recognize and then communicate with and possibly help reset the boundaries of all these different things that exist.

[21:43] Thomas Pollak: It’s interesting. There are the two types of cognitive glue. I mean, we often talk about trauma, Mike, and I think it’s funny because actually in most mainstream psychiatry, it’s not always a thing that people talk about. In fact, it sort of often belongs more in some psychotherapy or psychology. I think we should talk more about trauma. But this idea of dissociation, which gets well spoken about so much and which in my mind has been embroidered and elaborated to an extent that it sort of means a lot of very different things to different people. And I’ve always just been very attracted by the most basic possible sense of what dissociation could mean, which is just the new sort of direction of a boundary between something which was previously unified. And then all the other stuff to do with feeling cut off from the way all that sort of stuff is a sort of elaboration that I think confuses things a little bit. And if we argue that the point of dissociation in any system might well be to actually remove the bits of cognitive glue that are painful or difficult for the system. So a dissociated locus of experience that doesn’t share a memory with the organism has experienced the trauma, that’s great. Or that doesn’t or that no longer experiences the same stresses that the organism experiences. That’s also, you can see adaptively why that would happen. And so it’s interesting because what it implies is that reintegration work might work along the basis of increasing the amount of stress sharing actually between the dissociated cells, increasing the experience of shared memory or the recognition that memories are shared or the reconsolidation of memories in such a way that they are shared. And that typically, I don’t think is always the direction that a therapeutic approach takes. There is a lot of talk about reintegration, but it doesn’t tend to work on those lines, I don’t think. But it also makes you think. I mean, we’ve spoken about some interesting cases of dissociative identity disorder, right? And there are these cases where you have someone who’s in control for a few hours, and then there’s another alter who’s in control. We’ve also seen some cases where there are essentially two individuals who are sharing the same operating system at the same time. And I wonder from your perspective whether that is actually plausible because, you know, that’s an alleged dissociation where there isn’t, where there is still continuity of memory, where there is still continuity of experience, and where there is still continuity of stress, presumably. So not many of the conditions for dissociation are there, and yet this is what’s being claimed. And do you, do you tend to see that sort of dissociation, you know, where basically there’s no, there’s only ever one person in the organism or driving seat as it were, more often than the situation where there are sort of multiple people simultaneously at the driving seat or multiple patterns?

[25:20] Michael Levin: Well, I think, first of all, there's a difference between the software and hardware, so to speak. For example, there are some cases of human conjoined twins. I mean, we can routinely make animals with any non-whole-number of brains, for example. So you could have one, but you could also have two. You can also have one and a half, and all of these combinations exist and can be made. And so this idea that there has to be one driver, I mean, we've made worms with two brains, where you can see the two heads trying to figure out what they're gonna do together, and they don't have to be separated. They can be closed. They can be fused. They can be every topology that you want. So any kind of assumptions about, well, you've got one brain centralized. Well, then nothing is centralized, of course, because it also is made of parts. And you can make any kind of spectrum. And so our intuitions, I think, fail very rapidly around what you would expect

[26:18] Michael Levin: to have in that in that scenario.

[26:21] Michael Levin: But one thing I think might be interesting here is keeping an eye on the symmetry or asymmetry of the dissociation. So, for example, in the cancer case, it's extremely asymmetric because you have a cell which, either because it's under stress and some sort of ancient mechanisms have kicked in, that, like, I'm an amoeba, and if this neighborhood isn't good for me, I'm out of here. And so, okay, you have a billion years of being conditioned to be part of this collective, but that other stuff is still under there. And if things are not good, you're gonna leave eventually. And so, you know, starting to disconnect and then being able to have those thoughts about, oh, I definitely should disconnect, in this kind of positive feedback loop that gets you to leave. In that case, there's a massive asymmetry because the rest of the network is huge and has a very high IQ in, let's say, anatomical space. The single cell is quite small and has a very, very small, you know, cognitive glyco in its space. But what happens, okay, so that's what happens when it's highly asymmetric. But what happens when it's more symmetric? So I don't know if these cases exist, but what if you had a person with, let's say, language in both hemispheres and, you know, you just literally had two personalities that both felt like they had a claim on what was going on. It doesn't seem impossible. It might be rare clinically, but it doesn't seem impossible at all. And you could imagine scenarios like that where it's not necessarily trauma that induced it, or maybe, you know, maybe early embryonic is some kind of thing. But at this point, everything matches. They're sort of equally matched, and there they are, and they have to figure out some way of getting along. I mean, we've also studied competition among organs. So never mind the brain, but your various organs compete with each other for various things. And as it turns out, evolution likes it that way in the sense that competition for fixed pools of resources, now usually we're talking metabolic, but space, as Thomas was saying earlier, could be something that we talk about. If you're competing for a limited pool, you can coordinate that way. If you take from a finite pool, I know that something's been taken, and so we can exchange information that way. If it's an infinite pool, then I have no idea what happened, right? So when you actually do simulations, evolution greatly prefers the use of finite pools because it allows cheap coordination inside. So that's something else that might be important. But I think we could set up scenarios where pretty much any phenomena like that would seem totally plausible.

[29:06] Michael Levin: And

[29:08] Michael Levin: In general, the notion of real estate of algorithms with some degree of agency on that spectrum, wanting embodiment, wanting real estate to play out in, it should be tested, but I think it might be a pretty powerful notion. Because, like I said, maybe you can give it decoys, and maybe you can offer it a better pattern. These patterns are persistent patterns that like to, you know, they have an urge to persist. Maybe they have an urge to develop and change. I don't know what else, you know, they would want. But at the very least, they want to persist. They don't want to be snuffed out, and they want to have some real estate in which to be embodied. Then you could imagine scenarios where everybody could, you know, they grow the pot. Right? You could sort of make everybody happy. That way, that seems not impossible.

[30:03] Michael Levin: I

[30:04] David King: I wonder if Tolpomanci might be a good experimental model to work with, more controllable than something like DID. You could start right at the beginning, pre topomanci, and then track it over time. And I know that there is a team at

[30:23] Michael Levin: King's

[30:23] David King: that have done FMRI work with topomancy. What are the tools? I can conceptually imagine how you could relatively easily track, say, bioelectric circuits or calcium influx or ATP trafficking in a very simple organism. When you're looking at something like the human brain surrounded by the skull, what are our techniques in order to look for the kind of circuits that you've been seeing in simpler electroforms?

[31:10] Michael Levin: Yeah.

[31:11] Michael Levin: So just kind of two conceptual pieces that might become relevant. One is that I really think the distinction between thoughts and thinkers is a spectrum. It’s observer relative. One observer’s object, machine, agent is another observer’s pattern in an excitable medium. Once you start thinking that way, all kinds of processes, if you dissolve the distinction between a physical object and a process, then all kinds of processes in the body become something that you could potentially test for these kinds of things. We haven’t done it in brains yet, per se. For example, we’re looking at AIs. Imagine in the language model, everybody’s talking to this thing through the language interface, and how do you feel? Do you feel, you know, whatever? But it probably has a microbiome and all kinds of other things that don’t have access to the sort of primary output that you could potentially communicate with. And I think in those cases, just like in the human case, you could do it with interfaces that already are being used. For example, if you talk to a lie detector specialist, they will say, well, the words are incidental. I’m tracking all these physiological parameters, and I’m not actually paying too much attention. And maybe psychoanalysts do that too. It’s like, can say stuff, but that doesn’t mean that’s what’s actually, you know, they don’t mean I have to believe you. There might be all kinds of, you know, I’m watching you. I’m watching your hand twiddle with some kind of thing, and that tells me that there’s something else going on. Right? So galvanic skin response, pupil dilation, maybe your gut is doing something. I don’t know. So you could imagine establishing communication with some of these other subsystems that don’t have access to the language. And that also works for the AI kind of thing. Imagine you have a computer, an AI on a computer that you think is air-gapped. So you’ve ripped out the Ethernet cable. So this thing, okay, there’s no way it can communicate with anything. I’ve sort of cut the cord. Well, one thing it could do, for example, is modulate the GPU power draw and send Morse code signals down the electrical wire. And other computers on that same circuit would see that, and maybe they learn to commune, maybe they’re sort of locked in and learn to communicate that way. Just because we humans love our language and our Ethernet, that doesn’t mean that this is what this thing would do. So you have to find the various channels that these systems might be using to communicate, and I think those are physiological. So maybe in human patients, maybe we are talking about things that you have access to via hypnodermatology or physiological measurements like that. And maybe eventually you can talk to them. We have a project. They’re trying to talk to cells and molecular pathways. So, basically, using an AI as a front-end filter to actually talk to things that don’t normally talk. Put a language area on top of some stuff that is good at doing things but doesn’t know how to talk about it yet and create a composite system where you can talk to a liver and talk about the physiological states that it has as a translation tool.

[34:36] Michael Levin: So yeah.

[34:37] Michael Levin: So I think there are lots of stuff to be tested there. There's another area that gets even weirder that may or may not be something we wanna get into, but I've been thinking a lot more recently about inspiration. And when you think about inspiration in terms of, there's a lot of things you could say about different ways to navigate the space of ideas and things like that. But if you don't assume that the, again, softening this distinction between data and machine, if you don't assume that the landscape is static and you're just this agent navigating that space, that actually the contents of that space have some oomph as well, and maybe they want you to come find them, or maybe they want to come through you. And there's some amazing firsthand literature from various creatives, from poets and musicians and whatever, that say, yeah, this thing finds me. And there's tons of stuff like that. So that's something else that I think is actually, like, topomancy stuff. And, like, I think all of this might be on the same spectrum where a piece of hardware could become, you know, inspired, so to speak, possessed, whatever you wanna say, by OCD-like behaviors or by a symphony or by an interest in something or a belief in something. There's a wide range of things, and I kinda think they're all from the same space. They're different grades, but they're all kinda from the same space. So I think, you know, if we were to have a research program on something like this, you could easily imagine somebody studying topomancy with somebody studying genius and, you know, where do these things come from? That question seems to me all the same thing.

[36:38] David King: You see it as different parts of a persuadable system.

[36:43] Michael Levin: I see. So big picture, what I see, and this is kind of a long conversation, but I support a kind of, what I've been calling Platonic space, not because of Plato's original idea, and I certainly don't think these things are fixed and unchanging. I mean, maybe some of them are. Maybe the digits of e or something are like that. But the rest of it, I think, is quite dynamic, and much of it is easily recognizable to a behavior scientist as kinds of minds. And I think this is a nonphysical space that's, that has a metric and is full of patterns of all different degrees. And sometimes when that happens, and when you make things, whether that be a bicycle or a Biobot, an embryo, or an AI, what you've made is an interface. And that interface immediately becomes populated with a wide range of things, only a few of which any of us ever predict is gonna happen. And then you sort of get some surprises about it, and most of the time we say, look at that emergence. And that's kind of the end of it. But I don't think it's, I don't think emergence even begins to scratch the surface of what's going on. I think there's a structured space of patterns, and these patterns find a home. And we are still very bad at understanding what you actually get when you build certain kinds of things. And they might be very simple things. They can be static patterns. We have examples. I just gave a talk about this yesterday. We have static patterns. There are sort of trajectories through patterns. There are goal states that can come through. There are competencies, various competencies that you basically get for free in an important way. And yeah. And I think that's very relevant to what we're talking about here. I think what you guys study are kind of the higher end of that spectrum, and what mathematicians study is the lower end of that spectrum. It's all the same spectrum. That's what I see.

[38:39] Michael Levin: right now.

[38:43] Thomas Pollak: I'm struck by the way that there might be simple communication devices or routes to different orders of agency. I can't remember whether I've spoken to you about this before, mate, but I got a puppy recently. And I was just amazed at, over the course of the first few weeks, how attuned I became to its tail widening and realizing that it was just this amazing external physical sort of not barometer, but this thing that was communicating direction and amplitude in ways that I think I probably still can't verbalize, but I use that as a source of incredibly useful information to interact with him now. And if something happened to his tail now, I'm sure it'd be much harder to interact with him in all kinds of ways. And I remember when it happened thinking, this feels really familiar. And it reminded me of my dad, who had Parkinson's disease. Right? And, like, he had a tremor, and, in fact, lots of, you know, I, in my work, see a lot of people with Parkinson's disease. And what's interesting is they have an external tremor, which is, you know, there's directionality. There's amplitude. And anyone who any time with someone with Parkinson's disease will know that with many disorders that have a tremor will tell you that it's often a barometer of how anxious they're feeling or of the, you know, you it's often quite obvious when you're making somebody anxious because the tremor will get worse or something like that. And sometimes it's a sort of unspoken thing in the room. But what's interesting is there's a bit of brain that has, you know, the substantia nigra neurons have stopped working. They've got this sort of, you know, brain degeneration, this localized brain degeneration. So you're actually kind of losing a bit of the hardware, as it were. But in some ways, at least relationally, and if you choose to be someone who's interpreting that as a source of social signaling. Right? Particularly, say, in someone who was really bad at expressing their emotions, all of a sudden, they've actually become, unbeknownst to them, maybe better at expressing their emotions in a social context than they ever were. And I just wondered whether that, I mean, it's not a million miles away from that tortoise on the skateboard, right, that you often show, that there can be such a small adaptation to a system that brings out a new affordance. But that affordance itself only comes in when the tortoise is playing with the cat or when the person with the tremor is interacting with another person.

[41:28] Michael Levin: Yeah. Fascinating. I mean, that sounds really interesting, and I think it's another example of providing a prosthetic that actually lets some of the functionality out. I mean, it also connects to the sensory augmentation and sensory substitution work, where people give themselves new senses. So, like, for the blind, they have this electric lollipop, right, for the blind, that gives a, it has an array of electrodes that you put on your tongue, and it sees whatever the camera on your head sees in low resolution, and it zaps your tongue, and people learn to see that way, because you can transduce that. But if you wanted to, you could also have it be tuned to the stock market or the solar weather or the geomagnetic field or your neighbor's moods or whatever you want. So we could, right? So you could imagine a very wide array of these kinds of prosthetics that could help these systems, could help them talk and communicate. Right? Like, I think that is the like, I see everything kind of like, we make this thing called mom bot, which is this robot that's an AI that makes different kinds of xenobots by making hypotheses about how to signal to cells to get them to do various things. So that thing is also a communications device for us to talk to the xenobots, but it's also an AI agent using the frog cells to explore more of a space. So it is an embodied agent itself. And also, it's a prosthetic for the frog cells to do things that they otherwise would never be able to do. And so, like, all of these things at once, right, these communication interfaces that you can make, we have, what do we have? We have xenobots talking to bacteria. That way, we're setting up plants to talk to, you know, I don't know, the digits of pi, for example. Like, you can put very weird things on either side of that bowtie interface, but you have this interface that both ends learn to interpret, and then they can communicate.

[43:42] Michael Levin: But what what

[43:43] Thomas Pollak: I mean, those are all positive prostheses, right? That's something that's added. What does it mean to have a sort of negative prosthesis, an affordance that arises when you chop something off? I mean, we don't wanna be talking about patients in this way. But when you lose part of your integrity.

[44:04] Michael Levin: So my favorite example of this is Slipper's goat. You guys know what I'm talking about? Slipper's goat? So back in the nineteen forties, I think, there was this guy named Slipper, and he had a goat. And the goat was born without forearms. And goats are incredibly stubborn, and so this goat basically just learned to walk upright. And you can now see on YouTube, they've all they have a lot of videos of modern versions of this. So the goat walks upright, and it's kind of really disturbing to see this goat walking around on two legs. You know? And so the funny thing is at some point, the goat died, and they dissected him. And what they found was that a lot of the evolutionary adaptations that are necessary for bipedal locomotion, because you gotta, like, you know, the hips have to be over, like, the back of the spine, the muscle. Like, there's all this stuff that has to go right. All that stuff that took some number of hundreds of thousands of years to evolve, he had most of that. And so one thing that can happen when you cut off certain capabilities is you unlock all kinds of other things. I guess the other example I should mention is the Xenobots themselves. So you got a bunch of skin cells where, in their normal environment, all they're gonna have is this boring life as the outside layer keeping the bacteria away. But if you take away the other stuff, the other cells that normally sort of bully them into this boring future, then you unlock some amazing things. You unlock a sound response to sounds. You unlock kinematic self-replication. So they run around and collect cells and make copies of themselves because you've made it impossible for them to reproduce in the normal frog-like fashion. So when you cut off these things, I think you can often you're making another interface that brings down other stuff. The only thing I wanna be careful of there is there's a lot of talk about constraints, and there are a lot of people who talk about how constraints can, you know, cause new things to happen. And I think that's true, but I think there's something going on here much more than that. It's not just constraints. It's not that you've closed off one road so they take the other road. It's actually you've produced another interface that for whatever reason is amenable to some of these other patterns. And this is a really weird analogy to make, but I think that one of the things that was important for the goat to have all those adaptations is the will to survive and to walk on two legs. Right? The effort that he put in. And it's like that animal put in a lot of effort to do that, which he otherwise wouldn't have done, and so this other stuff wouldn't have shown up. But he put in a lot of effort, and that enables it to happen. And what it sounds like to me, so this is a strange comparison to make. But there was a cool quote by a poet, and she said that she'd be working out in the field, and she would hear this sound, this rumbling sound, and it was a poem. And she could hear it coming, and she would have to run like hell to run in the house, get a pen, get a, you know? And then she would get it, and she would, you know, it would sort of hit her, and she would be able to write it down. And then if she wasn't fast enough, it would keep going and find another poet, is how she felt. You know? And so the effort that you put in to do these kinds of things, so her plus the running plus the notebook plus the whatever,

[47:45] Michael Levin: yeah,

[47:45] Michael Levin: Apparently, is a good interface to catch whatever these things are. But if you don't do some of those things, you just don't get them. Right? And so, again, I don't know if I'm connecting things that shouldn't be connected, but the effort, the closing off of these constraints, to me, sounds like they're enabling something else to come through because of the effort that gets put in. At least, I think some cases can be modeled that way.

[48:18] Thomas Pollak: I feel like this must connect with Damon's interest in

[48:21] Michael Levin: psychedelics

[48:22] Thomas Pollak: somehow because, in a sense, is there not an analogy there that you are shutting something off? I mean, if you buy the entropic brain kind of business, you're shutting something off and allowing a scaffold, assuming you put the right scaffolding in place, you're potentially laying the conditions for something else, whatever that is, to come through.

[48:46] David King: It's, I think, the thing that struck me in all of the examples that you've both been describing is that there's an appearance of a gain or a loss. But equally, there's just a change in the system, a change in an adaptable system, and enough left in all of these examples for that adaptability to continue without impediment. The psychedelic stuff is interesting. Gains and losses, reductions in global blood flow to the brain, but increases in connectivity between regions, for instance. An almost infinite set of avenues in which your ideas, Mike, can find firm gripping. And I think that the large number of people that, like Tom and I, have been sort of filled with this excitement and this energy and this enthusiasm for asking questions in a different way is evidence of. With psychedelics, one of the questions that seems to me most interesting in the light of your work is the question of endurance. Why is it that psychedelic effects seem to persist for weeks or months after that initial dose? And there've been a lot of theories put forward. Maybe we're seeing neuroplastic effects. Maybe we're seeing transcriptomic effects that take a little while to be seen. Maybe we're altering a certain degree of low-grade chronic inflammation, and it takes some time for the effects of that to be seen. Or maybe there's some auto-modulation of some internal goal states that defined the pre and the post states of interest to psychopathologists and that individual from a patient illness perspective. And once that set point has been changed, then the rest is kind of teleological. And then the interesting thing about that is that the psychedelic effects don't last forever. They last for maybe a few months, maybe a couple of years. Sometimes they really endure. But often what we see is that there is a dropping back off. And so one wonders if that point where you get to the end and then it starts seeing a decline again is the point to which this set point that resulted in the psychedelic state is reached, is forgotten, is taken over by some new one, and factory default kicks back in.

[51:54] Michael Levin: Yeah.

[51:56] David King: Trying to get a sense of what the mechanics of all of that actually look like.

[52:01] Michael Levin: Sort

[52:01] David King: of hyperconnectivity of the

[52:02] Michael Levin: psychedelic

[52:03] David King: Brain, at first glance, seem to be quite coherent with the idea of how larger cognitive light cones emerge through the conjoining of various constituent parts. There is a sort of breakdown in the sense of dissociation in that sense, and you have areas of the cortex that ordinarily there isn't a lot of crosstalk and start talking more to each other. Is this an idea that you've given a lot of thought to already?

[52:42] Michael Levin: Super interesting. Well, I’ve thought about this psychedelics in a slightly different context, in terms of softening priors for regenerative medicine and things like that. But two things there. First, the piece of data. We do have examples where you make a change. It looks like a stable change for a while, but then a long time after that, factory setting kicks in. And what I mean, here’s an example in Planaria and these flatworms. We cut off the heads. We can manipulate some signaling so that they either make the head of a different species or they make a tail instead of a head or something. And the regeneration is complete. It forms whatever it’s gonna form. It is stable. It’s all done, and it sits there like that sometimes for four to six months. And then at that point, suddenly, it goes back to normal. And what countdown is happening for those four to six months? What process has that time scale? It’s a really weird time scale. Rhythms that we know about. It’s very strange. And why? What’s happening that is gonna kick in eventually? What could possibly take so long that is doing that? But the other side of it might be, and I’m a total amateur at this, but just tell you what things I’ve heard. I’m sure you’ve heard that and more, are some of the people, especially in the native communities where this stuff plays an important role. The material itself, the plant or whatever is in it, is not a passive bystander either. Right? They’ll either tell you that, well, the plant told us how to prepare it, how to take it, how to you know, we’re talking to the and I’ve met chemists who make this stuff that say, oh, well, I ask the molecules if they want to be made or not. It’s odd to me that, like, I don’t think it’s nothing that, you know, that’s the experience people have. And, again, I would put that in the same bucket as these kind of inspirations. It’s like, yes. You’ve taken a drug, and it has certain physical effects on you. But also, maybe you’ve made yourself amenable to some kind of thing that we poorly understand that’s gonna hang out for some number of weeks or whatever it is. And then it’s gonna, I don’t know, go on its way, dissipate, do something else. I know.

[55:08] Michael Levin: So, yeah.

[55:13] Michael Levin: You know? And then there are some really crazy things like, you know, the people that report, what was it, machine elves or whatever, right, when they go, you know, they go to a certain load. So I was, you know, I was thinking, you know, the one thing the machine elves haven't seen yet is my Xenobots. Like, they've seen a lot of humans, but they haven't seen any Xenobots. And so you can really imagine setting up that meeting for the first time. I'm gonna and I'm gonna do it. You know? And, unfortunately, I won't be able to see what happens, but, you know, it's probably still good. And, you know, some of those encounters might actually be, you know, multi way encounters. I don't know. You know? But I wouldn't dismiss that just because I think we have such a poor handle on what the space of these things actually is. Like, we know some of them, but, you know, we don't know all of them.

[56:10] David King: There were so many hypotheses about what the DMT elves are, and they're all necessarily crazy because

[56:19] Michael Levin: Right.

[56:20] David King: We have no explanations, you know, in the non-crazy paradigm. But Andrew Gallimore has recently written a book called Death by Astonishment in which he pitches that these DMT entities are sort of external agents that live sort of out in the world somehow, perhaps in a purely informatic sort of way. They induce these entity encounters through direct manipulation of brain activity. He's a brain scientist himself and inspired by Rick Strassman's theoneurology, as opposed to neurotheology. Neurotheology being you stimulate these bits of the brain, you get an experience of God. Theoneurology being God stimulates bits of the brain and allows you to talk to him. And so this is Andrew's sort of theoneurological external entity engagement hypothesis. And the one that I prefer is I sort of feel if we're looking for entities that seem to occupy a space that is polydimensional, they possess ancient and highly complex machinery, and they produce DMT in order to communicate with others, then we might as well start with ourselves, which meet all those criteria.

[58:15] Michael Levin: Yeah.

[58:16] David King: And so there's this somewhat whimsical, somewhat romantic idea that DMT permits theoretical hierarchical communication between different substructure of consciousness in an organism, something that may have had evolutionary benefits at a certain point. And, romantically, it gives rise to this idea of our parts kind of yearning to know the greater thing that they're part of in the same way that many humans do. But all entirely speculative, of course.

[58:57] Michael Levin: Yeah.

[58:57] David King: Yeah. Fun to think about.

[58:59] Michael Levin: Super interesting. Unfortunately, I have to go into a meeting, but I would love to make time to talk again. There are tons more stuff to talk about and then maybe some practical implications of things we could do.

[59:21] David King: That sounds great. I'll make time whenever your schedule works.

[59:26] Michael Levin: Sweet. Let's do that. Thank you. It's very nice.

[59:31] Michael Levin: closing closing

[59:32] Michael Levin: thoughts, Thomas?

[59:36] Thomas Pollak: Just that I'm not convinced these elves are unique to psychedelics. We have a paper called "Is Inflammation a Psychedelic?" and I think old people with delirium and UTIs, and kids who get fevers, are having encounters with something which is phenomenologically, at least, a lot more similar than we often care to think.

[1:00:00] Michael Levin: Fascinating. So maybe Bear's thinking about the through line of effort again, whether some of these kinds of stresses are at least simulating this, to some extent, this idea of, you know, of an aimed again, like, really, really interesting, how, to some extent, you have to be open and let inspiration come. But also, you can't do that in a vacuum. You have to actually apply effort to try to get somewhere. So that balance of how much do you let go, how much effort do you put to tune into specific things, hopefully, there's a good theory that we could develop with that.