Watch Episode Here
Listen to Episode Here
Show Notes
A meeting between Mark Solms, Chris Fields, and I: active inference, qualia, consciousness, brains, bacteria, and more.
CHAPTERS:
(00:00) Free energy, affect, midbrain
(11:42) Virtual governors and PAG
(19:24) Prioritizing stress responses
(29:31) Adaptive plasticity and agency
(34:44) Control-centered qualia and illusionism
(44:50) Global workspace, observers, emergence
PRODUCED BY:
SOCIAL LINKS:
Podcast Website: https://thoughtforms-life.aipodcast.ing
YouTube: https://www.youtube.com/channel/UC3pVafx6EZqXVI2V_Efu2uw
Apple Podcasts: https://podcasts.apple.com/us/podcast/thoughtforms-life/id1805908099
Spotify: https://open.spotify.com/show/7JCmtoeH53neYyZeOZ6ym5
Twitter: https://x.com/drmichaellevin
Blog: https://thoughtforms.life
The Levin Lab: https://drmichaellevin.org
Transcript
This transcript is automatically generated; we strive for accuracy, but errors in wording or speaker identification may occur. Please verify key details when needed.
[00:00] Chris Fields: I have a comment on your e-mail, Mark. We're completely on board; we're interested in the phenomenology of stress in arbitrary systems, down to bacteria, and treating physiological stress as some internal measure of variational free energy. The comment I wanted to make is based on some physics work about what's the structure of an agent that can deploy multiple interaction modalities, multiple ways of making measurements on the world and acting on the world. Mike, Jim Glazebrook, and I foresaw this in our 2001 paper in "Neuroscience of Consciousness" with the remark that interesting agents that had memory capability, et cetera, were going to be segmented into components that communicated classically. Jim and I have now generalized that: if you have a system that can deploy multiple reference frames that are distinct and don't commute, then those components of that agent have to be separable, so they communicate classically. What that says is that the different components of a system are measuring different components of variational free energy. Nothing is measuring the total variational free energy of the system, because there's no component that's looking at the entire boundary. Each component is looking at some piece of the boundary. Some of the boundary is devoted just to free energy exchange. That information flow doesn't get processed as information that one can do computing on. If we want to model the system's best guess about its own VFE value, we need to have some meta-processor that, instead of looking at the external boundary, is looking at the boundaries of the components that are looking at the external boundary and constructing a model of the VFE that those components are getting. That's the beginnings of an affect system. That's what the brain stem in effect is doing, or what the midbrain is doing.
[03:26] Mark Solms: That's almost exactly what I was going to say. This is the big issue that Mike brought to my attention. It never occurred to me that basal cognition descends below the brain stem, let alone below the nervous system, let alone to the level of bacteria. This has been the big eye-opener for me. Among other things about Mike's work, I became familiar with it through Mike. It was before I met you, Chris. Before going to the midbrain, I want to comment on what you said to start with there, Chris: to me, it's such an important part of the obvious that it bears emphasis. The system has to measure how well or badly it's doing in relation to these multiple component needs. They have to be treated as categorical variables, which is what you've just said. The importance of that for me is that it becomes a sort of principled basis for qualia, because such variables, not being reducible to a common denominator, must be distinguished qualitatively. That's terribly important for the whole business of consciousness and starting with affective consciousness, because we're talking about needs, need gradients. What is continuous? What is a continuous variable? I take your point that there's no component, at least at the level of the bacterium, and probably in any system that doesn't have this meta-homeostat, which is putatively in the midbrain. The common denominator is variational free energy, but there's nothing measuring it. There are two possibilities. One is whether anything has to measure global free energy. If each one of them is doing that, then the totality of the workings of the multiple components still has the same outcome, namely that free energy in general is being computed. It's just being computed in compartments. Because each compartment has the same imperative to minimize its free energy, there is a global effect of minimizing free energy. That's one way of doing it. The other way is the meta way, which we'll come to. I want to make clear that in my way of thinking, that common denominator is the valence. It is bad for any self-organizing system for its free energy to be increasing, and vice versa.
[07:33] Mark Solms: It is good for it to be minimized. For it to be reducing. There you have value, tied, of course, to survival value. Tied to the basic value system of all living things. That's the common feature of all affects: they have a goodness and a badness to them. They have a valence. But there's an additional feature necessitated by the very point that we are discussing, namely that there are multiple survival gradients, that in addition to valence there has to be a specific feel, a specific quality. I think it's terribly important that it is, at least as far as I can tell, obligatory, that there must be a quality. The value has to have a quality. There we're beginning to, as I said a moment ago, talk about conscious feelings, or at least conscious proto-feelings or proto-conscious feelings or something of that kind. That's dramatically important. Now to come to the other route, which is the meta-homeostat, I think you have exactly in mind what I have in mind when you say that we need to go to the midbrain: there is this terribly interesting fact that all of these homeostatic, all of these need-monitoring systems send their residual error signal to the periaqueductal gray. The periaqueductal gray, in those of us who've got such a thing, is therefore — I don't know whether we can think of it as computing variational free energy in totality — it appears all that's necessary is a prioritization of those different homeostatic deviations. Which one of these, to the extent that there's an action bottleneck — where there's not an action bottleneck, where it's autonomic, it doesn't matter. But where some sort of action in the external world is commanded, then we start to come across action bottlenecks. Which one of these needs am I going to prioritize? Which one am I going to act upon first in the world? It seems to me that's the main thing that the PAG is doing: it's prioritizing. In the vertebrate PAG, immediately dorsal to it, is the tectum with its superior colliculi, which is getting a condensed mapping of the multisensory environment, what some people call a saliency map or an intentional map. So between PAG and superior colliculi, the interaction between them provides the possibility of evaluating needs in relation to opportunities, the colliculi being the map of opportunities and the PAG being the monitor of needs. That is a splendid idea, which is not my own. It was first formulated by Bjorn Merker, a Swedish affective neuroscientist, who is retired now, living in a forest somewhere. That's my way of thinking about the issue that you've just raised.
[11:42] Chris Fields: How far down in phylogeny that kind of metaprocessing system goes, that keeps this kind of multidimensional representation of stress updated, is a good question. In something like E. coli, you might just have a bunch of stress components that nothing is trying to integrate.
[12:11] Mark Solms: Yes.
[12:12] Chris Fields: Probably has a better understanding of that from a physiological point of view than I do.
[12:21] Michael Levin: I wonder, Chris, when you first started talking, you said down to bacteria. What about all the stuff below that, specifically in the case of stress, issues of geometric frustration and materials, the things people see in material science? On the one hand, you might think of magnetic domains trying to align. You might think that's what you just said: it's a bunch of stress with nothing integrating anything. That may be. I also wonder: there's this notion I see referenced from time to time, and I've had a really hard time finding the originals, but it's Wiener's notion of the "virtual governor." It's the idea that in certain systems of multiple linked components the whole thing is rather unstable, and trying to control it by controlling each one is really hard. What he was saying is that you can define this global thing. In his case it was a bunch of dynamos connected together and they were super unstable that way. What you can do is define this virtual super-dynamo that doesn't actually exist, and formulate a control strategy for that thing, which works better than trying to manipulate the actual stuff that's there. I wonder if in certain systems that may not be on the phylogenetic tree at all, that may be far simpler, there isn't some non-obvious virtual integration center doing a very primitive version of what you were just talking about: trying to figure out which stress to relieve first. Maybe this is a question for somebody like Robert Batterman, who studies mesoscale materials: you put a material — maybe a dumb material, or maybe an active matter system — under stress. I wonder if there's a good lens to say here's a virtual thing deciding which buckles first and which stresses to relieve first. You could even imagine a multimodal system with physical pressure, heat, and magnetic domains — a multisensory thing. I don't know if this goes further down. I don't know that anybody's done this, but I can imagine defining such a thing.
[14:52] Chris Fields: You can certainly think of something like this on a skeleton as an elastic stress detector that has its own response capabilities built in, but also is signaling constantly to the rest of the cell about elastic stress or mechanical stress.
[15:14] Michael Levin: Gunther Buehler was saying that the microtubule organizing center was this sort of integration module for the cytoskeleton, trying to make decisions about priorities and the morphological computations the cell has to do next. He had a lot of prescient stuff in the early 90s. I'm talking about the intelligence of cells implemented by the cytoskeletal network that's inside. Mark, I wanted to ask about this special region of the brain that you're talking about that does this. Is there anything special about it aside from its connectivity? In other words, if you saw it in isolation from the rest of the brain, would you be able to say that this is the kind of thing that would do that? Is there anything special about the neurons? Or could any collection of neurons do that if suitably connected to the rest of the brain?
[16:27] Mark Solms: At the neuronal level, no, it doesn't have any special type of neuron. It's a rather poorly differentiated structure. In all vertebrates, as the name suggests, the periaqueductal grey is around the cerebral aqueduct of Sylvius, which is the canal, the central canal, which runs through the neuraxis, up the spinal cord, right up into the hemispheres. In the hemispheres, it's the lateral ventricles, and then there's the third ventricle between the two thalami, and then there's a fourth ventricle between the pons and the cerebellum. Between the third and the fourth ventricles, there's this little canal. Behind that canal is the tectum, which is, as the name suggests, the roof over the canal. The periaqueductal grey wraps around the rest of the canal and terminates on the tectum. It's 14 millimetres long in humans. It has a columnar structure, which, as I say, is not very well differentiated. The columns themselves are not that well differentiated from the surrounding white matter. The lack of differentiation partly speaks to its age. It's an ancient structure and its location around the canal, the central canal, is a precursor of what will happen higher up. At the floor of the third ventricle, the hypothalamus is there. There are several what are called circumventricular organs. What that location bestows on them is the capacity to sample CSF. They are sampling; there's blood flow through them. They're sampling what's going on in the blood, they're sampling what's going on in the CSF, and then they're getting these neuronal messages from all of these various lead detector, other distributed lead detector mechanisms.
[19:24] Michael Levin: That's really interesting. This is the sort of thing I thought a long time ago when I first heard about the so-called pleasure center. This, the currency of reward. I thought, would you know that's what it was if it was sitting out in a dish somewhere? Is there something special about it or could anything do it? It's just, it has this functional connection with the rest of it. What does a pleasure center do all by itself, if anything?
[19:54] Mark Solms: In the spirit of all three of our work, I would say it all comes down to what kind of computations are being performed. It doesn't matter what the substrate is that's performing them. Doubtlessly there's more to be said on the topic of the finer structure of the PAG that I wouldn't be able to speak to because what I've told you are the limits of my own knowledge about its structure, but there may be more. For the reason I've just said, I doubt that that's going to be the decision. It's more, it's what sort of information it's processing and in what way it's processing that information. Now, at the level of E. coli, let's pause there and let me ask: these various components of stress that need to be monitored and acted upon, do the actions have to be prioritized, or can they be performed simultaneously? I'll tell you why I asked the question. It's because it's coupled to a second question. So let me ask the second question first and then tell you why I asked both questions. The second question is: the actions that it can perform in response to deviations from its preferred states — are those actions monotonous? In other words, is there only a stereotyped response available to the organism? I think it's a fundamentally important point that if these are perforce categorical variables, they are in their very essence qualitatively distinguishable from each other. I said that is a principled basis for why qualia would come into existence.
[22:43] Mark Solms: With a valence, remember, these are qualia about which the organism gives a damn. Existential value they have. But if in addition to that you have to prioritize them, then there has to be each of these may be categorically distinct, but if they're being processed simultaneously and independently of each other, which they can be if there's no need to prioritize, then there's a less pressing need for them to be qualitatively distinctive in the sense of experienced qualities. Because if you're trying to juggle balls and you're wanting to know which one you want to catch next, there needs to be some way of identifying it. So it would need to be color-coded or flavored. And the second reason, implied in my second question, is that you would then have to feel your way through the success or failure of your action if the action is not just a monotonous reflex but has a trial-and-error component. "Let me try this, let me try that. Is this working, is that working?" obliges us to assume there's a palpating going on. That starts to suggest something in the nature of, if you're palpating multiple systems, you need to distinguish them from each other, and there needs to be some particular way of representing how well or badly it's going categorically. I just don't see how it can be done without it being done by means of qualia of some kind. To say again the obvious: this is so important because David Chalmers and his predecessors, like Tom Nagel and others, claim that there is no possibility of a mechanistic account of why there is something it is like to be the organism. They say a mechanistic account of "what it is like," in other words a subjective, qualitative, phenomenal experience, cannot be accounted for mechanistically. I think that what we are talking about right here and now is a mechanistic account of how qualia come into existence. As long as you take the step, which I think is absolutely essential, of contemplating the mechanism from the point of view of the organism, I don't see how we can have a science of the mind without it. You have to take the observational perspective of the organism.
[25:34] Chris Fields: I'll say at least two things about priority. One is that all of these response modalities require energy. The system has finite energy resources at any given moment, and many stressors decrease the energy supply. The system in general is just not going to have the free energy available to deal with everything all at once. That's one point to do with energetics. The other point is that a lot of these responses require gene expression. You have the question of how many, in something like a bacterium, operons can be transcribed simultaneously. Operons overlap in various ways. The polymerase for one is going to get in the way of the others. So there's not just an energetic issue. There's also a structural issue to do with the way the genome is organized that will prevent the system from responding in every conceivable way at the same time. I don't know the details of E. coli genomics in terms of where relevant stress response operons are. But for example, if your stressor is running out of the sugar that you're currently equipped to process, then you need to express the genes for some other sugar. If those operons are nearby, then you have to turn one thing off to turn something else on. In mammalian or eukaryotic systems where the signaling pathways are so interwoven and cross-modulatory, there are all kinds of winner-take-all or winner-take-most arrangements between signaling pathways, which turn one off when another becomes active. This may be an example of what Mike was describing in terms of a virtual governor, that it's just the way things are arranged to enforce certain kinds of prioritization.
[28:35] Mark Solms: That's just fascinating. And what about the repertoire of responses? Are they fixed and immutable or are there trial and error possibilities? In other words, the possibility of something that I would describe as voluntary action.
[28:56] Chris Fields: At some outer limit, they're fixed by the genome. You can't make proteins that you don't have genes for. Even there, you have the trial and error of gene transfer. You can go out in the community and try to get new genes to do something that you need to do. Whether that can be treated as voluntary or not, I don't know; at the level of microbes, for example, they certainly do it all the time.
[29:31] Michael Levin: I want to focus on two things. One is to broaden out. If we think beyond bacteria, you can't make proteins that you don't have genes for, but you can take actions in physiological and anatomical space that you don't have genes for. You don't have genes for anything, but in those spaces. There's huge room for novelty in problem solving in those higher spaces. Maybe not in the protein space, that's true, but in these other spaces there is. In particular, I'm thinking of a related model. We're just starting a project now that will give a lot of data on these things, which has to do with our barium worm example. What happens there is you throw barium on these planaria. Barium is a non-specific potassium channel blocker. The cells are very unhappy, especially the cells in the head because there's lots of neurons. They want to have potassium channel function. Their heads explode physically. Just overnight, they blow up. Then the next couple of weeks, these worms grow new heads that are completely barium insensitive. They don't care. Several years ago we asked a very simple question: what's the difference transcriptionally between the barium-adapted heads and the normal heads? We found that there's a pretty small number of genes—only a couple dozen—that were differentially regulated in response to this stressor. Planaria never see barium in the wild. There's never been selection pressure to know what to do when you're hit with barium. That raises lots of interesting questions because it's improvising a transcriptional solution to a physiological stressor that it's never seen before. Is it generalizing from things it had seen before, which may be epileptic hyperexcitability? One thing we're going to do is try to track cells, worms, and even groups of worms through this journey to find out how they explore that space, because it's like a 20,000-dimensional space of possible gene expressions. You don't have time for random search. We don't know whether it's a directed walk to the right region or whether they're ambling about and some of them find it. Do they all take the same walks? But a lot of the changes you make will kill you long before you get anywhere good. You can't just randomly start turning knobs on this thing. That raises the question of creative problem solving, generalization, and navigation in this weird space where you'd like to be but don't know how to get there. I don't know if they have an internal coarse-grain model of their own, that mapping between the transcriptome and the physiome.
[32:06] Michael Levin: which genes do I now regulate to have a particular physiological capability? Do they have a coarse-grained manual to their own cell function somehow that they know roughly what to do? Or is it a completely lightly directed search that they do? I think all of these kinds of notions of where do you go, what do you pay attention to, it's all going to play out there. We'll see how that works out. Another thing I wanted to say: I wonder if there's a name for this; maybe you guys know. If there isn't, I think we should come up with a name for it, which is the flip side of qualia. People who are interested in consciousness talk a lot about qualia. That emphasizes the receiving end, the inputs. What does it feel like to be on the receiving end of various things that happen? I'm really interested in the opposite side, which is the active end: what does it feel like to try to figure out what can I do next? What effectors do I have? What should I do next? It's the output end, which to me is really the critical part of the whole consciousness thing. Yes, what does it feel like to be on the receiving end of being a sentient being in the physical world? But what's the flip side of that, which is I have to take action. What do I have control over? What do I know? What parts do I have? What are my effectors? Where is the boundary between me and the outside world? Do I really control over there? These are the effectors that I have. That active aspect—the fact that you can't, even if you don't believe in free will, sit back and see what the Big Bang has ordained for you; there's no way to live that way. You have to be active all the time. Is there a name for that? Like the flip side Aqualia.
[34:44] Mark Solms: I disagree that the qualia are on the receptor or are particularly attached to the receptors as opposed to the effectors. I think the qualia are attached to the control center. I know that you guys are very familiar with these sorts of things. The sensory states that are of interest are actively authored sensory states. In other words, I have acted on the world, given my model of the world or my body, I have acted on it. I expect, given that model, to have this sensory consequence flowing from my action. If the sensory consequence is what was expected, then all is hunky-dory. If it is not what was expected, then you have an error signal. The average error is what we need to minimize. Within one of these homeostats, that is just the state of me. It's not the state of the world. It's the state of my free energy. Given the point you've just made, Mike, about it being an action with expected sensory consequences, then what the control center is monitoring is changes in expected free energy. It's even more action oriented when you start speaking about expected free energy as opposed to free energy. This is, in a way, the point I was making with Carl yesterday; he made exactly the remark you just made, saying, well, consciousness is all about action. In the discussion we had yesterday about illusionism versus realism, this point came up. I noticed that some people in the meeting, some colleagues on this panel, seemed to think what you initially said, which I'm disagreeing with. I can see from your nods that you're not going to defend the view that qualia have to do with what the receptor samples.
[37:49] Mark Solms: And that seemed to be their view, that the whole idea of the illusion, this is what they were saying, the illusionists in the room were saying, so what we are, our qualia, what we are sampling, our sensory states are not the world. They are our sensory states. They're states of our blanket. And they are sampling just small components of reality, the ones that are of survival interest to me, and they're representing them in any old way that I like, and this is what that Chuck Donald Hoffman calls icons. It's just like then it speaks of a user interface. I accept all of that, except I don't think those are the qualia. I think that the qualia are my response in terms of what matters to me, that this active aspect, again, that you're speaking to. And that's an internal state. It's me measuring my free energy or my expected free energy. Are things turning out as expected or is uncertainty prevailing from my current actions in relation to sensory parameters, which are not in and of themselves the point? Because remember, as we get increasing complexity, and I don't mean of the predictive model so much as of the creature, the homeostats are no longer tied in a one-to-one fashion to a particular sensory gradient or, for example, an affect like fear, which is pretty basic. What sense is it? What sensory modality drives fear? A whole lot of them. So it's what the homeostat is having to do. The control center is doing a pretty complicated job there. It's busy deciding, as it were, is this constellation of sensory inputs or error signals a fear constellation? And maybe fear is too complex an example. You could say pain, likewise. It's not just one sensory gradient, and it's a whole lot of things.
[40:55] Michael Levin: Yeah, I think it's.
[40:58] Mark Solms: Just to round off the point, since we've ended up talking about illusionism. My point to these colleagues yesterday was that qualia are not an illusion. It is the direct registering of your own free energy. In other words, it's a measuring by the system of its internal state, its response, its subjective response to the actively altered sensory inputs or the deviations from the expected sensory inputs. So you can't call that an illusion. In what way can that be an illusion? It's not about some world beyond the blanket. It's about yourself. What could be more obvious than that our qualitative conscious states, or subjective states, are states of us; they're not states of the world. Illusionists don't disagree with that, but when it comes to affect, which they don't think about, I think affect is the foundational form of consciousness for all sorts of reasons. When consciousness emerged in nature, and through you guys I'm learning it emerged a lot earlier than I ever thought, it would have taken the form of the kind of thing we're talking about now. It then becomes a whole different story once an organism is able not only to feel its own state, and remember why it has to — we've been discussing, in the earlier part of our meeting today, all of those imperatives — why it must be a qualitatively differentiated subjective state of existential consequences. That is the subjective and the valence aspect of it. So that once you have these "I feel like this" states — if I can put it into words; such small organisms are not thinking in words — it's just the raw feel. Then the additional adaptation becomes "I feel like this about that." Then we extend our affective fields onto a representation of the world, which must have some common currency between the feeling and what it's about. I think that's what perceptual consciousness then becomes. It's a sort of application of values onto the context within which that feeling is generated. You can imagine all the functionality that flows from that. Now you have the possibility of working memory, of saying, "I feel like this about that," so I must avoid that and I must approach this. All those things start to become possible.
[44:50] Michael Levin: I think you're 100% correct. I was just saying that I think, as you said too, a lot of other people focus on the input side of qualia. And one example of this that I think is interesting, and it's why I'm looking for a new terminology, is in all of these theories of consciousness. So we have epiphenomenalism. So the point of epiphenomenalism is that there is somebody home that feels things happen, but they have no causal power. They don't actually do anything. But what we don't have is the flip, to my knowledge, nobody's proposed the flip kind of theory that, yes, there's action, but nobody actually feels what it's like. So people clearly prefer; they've clearly emphasized the input side of things. But nobody's bothered to. Nobody's chopped it off in the other direction that I know of. And so that says to me that going all the way back to when these things were first proposed, there's this emphasis on the receiving end as opposed to the active end.
[46:04] Mark Solms: Well, I get back to human luck.
[46:08] Chris Fields: You're both pointing to the fact that this language was invented by people who viewed perception as a passive affair.
[46:19] Mark Solms: Yeah, exactly.
[46:20] Chris Fields: And had this idea of the observer that runs throughout classical physics, the whole classical worldview that the observer's not doing, and it doesn't have to do anything except just sit there and collect information. That's what the inactivists are rebelling against in their confused language.
[46:50] Mark Solms: Yes, that's right. That takes us to Helmholtz, the whole thing of unconscious inference. That was his point: what is perceived is what is actively sampled. Carl always acknowledges that his whole way of thinking goes back to that early work of Helmholtz's, exactly on this issue of the active nature of perception, that perception is the consequence of an action. I had a slightly wild thought while I was listening to what you were saying, Mike — not when you asked this question now about the flip side of qualia, but when you were answering the question about the extent to which we may describe the action repertoire of, say, a bacterium. You talked about these worms that you study as including voluntary actions. And your answer is extremely interesting. As I was listening to that, I was thinking about the question that Chris started us off with in our conversation today about whether there is anything that is orchestrating all of the different components. The wild thought that suddenly occurred to me is that what's orchestrating it is consciousness. It's that when you start thinking as we are in our conversation today, and you take this route and then that route, ostensibly discussing different issues, they all just converge on the same answer. It seems as if something like consciousness has to come into play if the system is going to be dealing with these problems. So it just occurred to me that basically what we're saying is what they nowadays call, in relation to much higher, much more complex creatures, the so-called global workspace. That's the whole idea. So that suggests that the overarching function is the global workspace. And the global workspace, therefore, is not an epiphenomenon. It's doing something which wouldn't be possible otherwise, and if that is what's doing the orchestrating, maybe that's the virtual regulator.
[50:03] Michael Levin: I've got a student with whom we are running through the various popular theories of consciousness, half a dozen or so, and just asking whether, for the same reasons that each of those theories associates consciousness with brains, we should then have to posit the same thing in the rest of the body, because many of those things—people talk about cytoskeleton, magnetic fields, this action-perception cycle and orchestration. These things go on in every part of the body all the time. It's actually very tough, on any of these theories, to be able to say, oh, well, clearly it's the brain and it can't be this other. If, as I think about it, we can, and we know very little about how it works, you can imagine that with this trying to navigate your way through this physiological and transcriptional space with disastrous consequences if you go the wrong way, is it something to be part of that process? It seems like it would have to be. It seems to me it would have to be. And all of that, at the same time, is also something to be that worm in whose body—those are not the same. It seems perfectly plausible that there are multiple distinct consciousnesses associated with that system doing their various jobs in various spaces.
[51:37] Mark Solms: I was reading a paper today that you wrote on nervous system–immune system interactions. I can't resist making a further leap based on what you were saying, Chris, about the idea that the observer sits there and observes. That point that you made there is what leads us in physics into a participatory interpretation of the whole of quantum mechanics; the whole problem of the observer arises from that. It's not a problem that only psychology and neuroscience have carried. The passive observer of Newtonian physics gets problematized, and you can't leave the observer out of it all.
[52:58] Chris Fields: Looking back, what I think was really happening was the observer was being treated as a god, as not being a physical system at all. If you treat the observer as a physical system, then you're immediately faced with Newton's third law, which says interaction is interaction. You have to be inactive. You have to be talking about the action side as well. So the only way to really have a passive observer is to have a supernatural observer.
[53:38] Mark Solms: Yes, God's eye point of view is what you're talking about.
[53:45] Michael Levin: There's a really interesting project that I've got another student working on now that's with a very simple — I'll show you guys once the results are ready. A very simple system, which are sorting algorithms. You have a string of numbers, they're randomly distributed, and there's selection sort and then bubble sort. The whole algorithm is maybe ten lines of code tops. But of course, they also always presume two things. They presume an omniscient God's-eye observer, which can see all of the pieces and moves them according to his plan. It also assumes infallible hardware: when you want two things to swap, they swap. We released both of those constraints and made versions of these algorithms that are cells' eye-view algorithms that talk about what each cell wants: neighbors in a particular order, but all the cells want that. When you do this, some amazingly interesting and bizarre capabilities of these algorithms show up that are nowhere in the algorithm itself. They're not apparent from looking at the actual algorithm. They do interesting things because you can play all kinds of games. You can make different cells obey different algorithms. They can be a chimera of two different sorting algorithms, and you can ask how the two interact. And how they navigate that problem space of going from a jumble to a perfectly ordered sequence: they all have a target of perfectly ordered. They do all this. This, I think, goes back to what Chris was saying: you can't make proteins you don't have genes for, but apparently you can have functionality that you don't have algorithmic steps for, because the simple version of this is just emergent complexity, but I think it's more than that. I think there's more hiding in these things that you can gain access to that are in no way apparent from the examination of the details. That observer thing is, I think, really interesting, moving it around, and also the evolutionary aspect where it...
