Joscha Bach: Could Xenobots could be Conscious?

[00:00:00]I was speaking to Josh Bongard yesterday who is Mike Leven's collaborator in the creation of these xenobots. You know, these are cell aggregates of heart and skin cells from the xenopus levus frog.

[00:00:15]And what they did was, you know, there's no neurons, no brain cells, but what they did is they started modeling the calcium spiking patterns within these cells. And they realized that those patterns when they really looked at at what the the structure of those patterns was, it was completely analogous to neuronal firing patterns in the brain.

[00:00:36]And then we were thinking like, well, what what what could be encoded in those calcium signals in this cell aggregate and it seems to be doing all of these different learning and memory, has differential gene expression. It has it does kinematic replication which I know is controversial and some people will say no that's not replication because there's no gene transport transfer but it's doing all these weird strange agential um you know behaviors properties features and seems to me like that thing that cell aggregate has has experience now that's something between you know a living biological organism and something that isn't Do you do you think that the xenobots I'm I'm sure I'm I'm sure you're familiar with this work.

[00:01:23]Do do you think that a xenobot would be conscious? Would you would you ascribe second order perception to this lump of cells?

[00:01:33]>> No. So I'm not I'm not uh ready to rule it out. But my default uh opinion was would be that first of all I I don't really know but uh is it actually necessary for producing such behavior to have that? And uh I know from uh that we can build things in computers that are producing very complicated patterns very easily. For instance, there's this artificial life community that people should look into. It's really beautiful.

[00:01:59]And uh they start out with all sorts of automata in initial thing where uh this game of life uh it's very simple um cellular automaton which uses a ukian grid. So basically just little boxes in uh in a rectangular alignment and uh each of these boxes can contain a pixel or no pixel. So when it's there's a pixel in it, it's alive and when there it's empty, it's dead. And when a cell has exactly um three neighbors, it uh comes alive. And uh when it has uh more than four neighbors, uh it dies because it suffocates. that have it less than three neighbors. Uh it dies because it's too lonely. And as a result, when we start with an initial pattern, just execute the simple function, we see an evolving pattern.

[00:02:47]>> And uh you can build stuff that looks a bit like organisms. You can build little gliders that are replicating themselves and moving around. You can even build computers from them. And this still looks very mechanical because it's constrained to this very simple grid and it's just binary. And people extended this into a paradigm called linear. And the linear are moving in a continuous space. They can move in every direction.

[00:03:10]And uh they can have more or less continuous states in the cells uh of the computation. And as a result, you can build stuff that really looks like little organisms uh that you would see under your microscope like little cells moving around and performing very complicated dances. And uh this looks really alive. That's why it's also called this artificial life community.

[00:03:31]But in in order to behave like this this uh you just need to observe a set of very simple rules and just by the application of these simple mindless rules you get emerging complex organically looking behavior similar to example is swarm behavior in birds for uh to form these complicated swarm patterns. It's not necessary that the swarm has a might of its own or that the individual birds are modeling the shape of the swarm. Instead, we can get the swarm behavior by letting every bird or every fish just follow a handful of very very simple rules that only let them react mechanically, so to speak, to their what they observe in their environment, what the other animals around them are doing. And as a result, you get this emergent pattern in their behavior. And so uh this let's lead leads me to think that I should be hesitant to ascribe uh intelligence or sentience or feeling or intentionality to uh everything that looks like it was complicated.

[00:04:32]>> Yes, I understand that.

[00:04:33]>> So so have all these mechanisms that are the result of evolution that leads to career and behavior when the arrangement is correct. But it's not clear if the xenobots are uh sustainable individuals, right? There probably are not. You need to create very specific environment so they can live out their days. And uh it's I think a lot of the stuff that you observe there is ephemeral. But uh the basic idea that Michael Leven has is that organisms are less clearcut as we make them out to be is probably very sound. It's very inspiring. So this uh idea that an organism is more like a state uh of humans like a country and uh then it is like a single thing and if a country dies is not like all the individual die. Many individuals will die because they die of famine but many also will try to move into the woods and survive and do something for as long as possible. And so maybe this is what Mike Leven is observing with the ginobot. So if you isolate a few cells from uh the skin of the frog, they will try to make do and do as best as they can to uh uh build some local structure to survive.

[00:05:43]Not all uh creatures do this like when you take an ant from its nest, it uh gives up at some point and uh because it has no purpose anymore.

[00:05:54]>> Yeah.

[00:05:54]>> And maybe this is also when you take humans away from their mission, many of them give up.

[00:06:00]It seems tricky to me. So, you're saying that if we accept that a frog is conscious, but 200 of its skin and heart cells aren't, there's some line in the sand between a frog and a clump of its cells that consciousness appears or emerges. That's that's what's tricky to me. I can I can completely understand that you can get very complex patterns that seem on the surface to be conscious, but there's no reason to believe it is. But then on the other side, it's very hard for me to to draw the line in the sand. Like where's the magic sauce? Where does it emerge? If there's somewhere between a frog and a clump of frog cells is a continuum of conscious to non-concious, like where is that line? I'm curious to hear your thoughts on that.

[00:06:46]>> A lot of people think that conscious is is somehow on a continuum with different degrees on a continuum. And I don't quite observe this in myself. To me, consciousness seems to be binary. It can be lucid to different degrees. Uh abundance pattern is either sparking itself uh into existence. It's igniting or it's not. And it's really like an ignition when you wake up in the morning or uh when you come to in a dream uh that you realize uh that consciousness is there. And if it's not there, you can be a sleepwalker even uh right it can be mindlessly uh playing out automatic programs that you have acquired uh while you were conscious.

[00:07:27]But uh I'm either conscious or or I'm not. It's almost like pregnancy. It uh there is a gray area between being pregnant and non-preg. But it's not a stable interesting rail where interesting stuff is happening. It's simply the face transition. And so there is a phase transition while you space out while you fall asleep and go into deep sleep or a face transition while you wake up from anesthesia or from deep sleep. Uh but uh you're either there or not and you you can observe this in yourself, right? what it means to space out and lose consciousness and to fall asleep and you there's still many functions that you can perform but not you because you're only there when you're conscious that your organism can perform that your even your brain can perform while you're not conscious but even coordinated action is possible to some degree but it's not coherent there is nobody home who is able to react and relate everything to a single vantage point and so I suspect that consciousness is necessary uh when You want to relate everything to this single vantage point and create this coherent sense of now and agency and for the other things it might simply be not necessary.

[00:08:37]So I understand that you you on a phenomenal consciousness level you either are conscious or you're not. But there's also likely as we've said many different flavors of consciousness as in a dog dog consciousness is different cat consciousness octopus consciousness. So, do you think >> you think that dogs and cats simply have a different personality?

[00:08:59]Their consciousness might be similar.

[00:09:01]>> Do you think it's it's useful to compare those levels of consciousness, the quantity of consciousness within a human compared to the quantity of consciousness within a dog? Is that a useful comparison?

[00:09:13]>> Yeah, I think if uh we take this uh my attempt as a definition or characterization of consciousness from the beginning of the podcast seriously that it's a second order perception.

[00:09:22]It's uh not quantitative. If it's qualitative, you either perceive that you're perceiving or you don't perceive that you're perceiving. And uh so this waking up in the morning is that something notices that noticing is going on. Right? The second order perception is happening and it feels like an ignition to me. It's really that once this is fired up, it's able to make sure that the perception is continuing to go on and continue to make sense. And if this is if you're spacing out again, there is no guarantee that your actions are continue to make sense. that everything falls apart and fragments and dissociates and dissipates. So it it seems that consciousness is the simple solution for creating a mind of a type like ours. And there is of course a difference in in cats and dogs and humans. But uh it seems to be for instance in uh a difference between cats and dogs is to which degree they are self-interested or their reaction times are very different. Their aesthetics are very different. The way in which they relate to themselves in the world are very different. Cats are much more ornamental than dogs are for the most part, right? And much neater and uh much more self-interested. Whereas dogs try to find somebody to latch on to and less self-directed than cats are. But uh I think that they perceive themselves to be perceiving and they can also do this for you. So basically the dog notices that you are noticing that you are noticing. So the dog is aware of the fact that you are unconscious and vice versa. So you can create this feedback across organisms where you share conscious states where you basically have states together where you build real-time feedback real-time resonance and create a shared now which is one of the interesting features of consciousness that it's necessary and uh facilitated for perceptual empathy where you basically build shared mental states and this is also a reason why it's I think uh quite undeniable that cats have to be conscious because when you have a shared mental date with the cat, right?

[00:11:23]How would that work if the cat was not conscious?

[00:11:26]>> Yes. No, I I agree with you there. But what I'm I'm still really curious on this this categorical distinction between the the the xenopus frog and the 2,000 xenopus frog cells where one is conscious and one isn't. There has to be like if you >> So, so what's the difference? is >> uh if you look at a government for instance, if you put a gang together, they're not necessarily a government because it doesn't necessarily scale, right? Even in if the individuals build binary relationships between each other, uh it doesn't scale. And if they build a reputation system across each other, it doesn't scale beyond the tribe because it means that everybody has to keep the books on everybody else, right? And gossiping only gets you so far. And so a tribe is limited in its coordination.

[00:12:14]But to build something that makes an unlimited number of people do the same thing like agriculture requires the specialization of across many different groups of people that need to coordinate. You need to give them a shared spirit. And so maybe consciousness is necessary to impose a shared spirit, a shared scalable agentic software on very large set of sub agents and uh under which conditions this feels like something I don't really know. That is something that we I also not saying that the casino is not conscious. I I simply don't know. uh and then when I ask myself what would be the evidence for this it would have to be that if I actually understand what it represents and uh because I cannot talk to it and I cannot have a shared mental state with it that I can recognize I would need to be uh able to actually make it talk right and I can only do this by looking into the representations that it builds inside of a simulation of the same behavior so I don't see a good way to to resolve this uh unless as uh there is some kind of wild guys that has already made these experiments and then is then talking through Mike Leaven or Steuart Hammerov or all these other modern mystics that uh feel that they have a greater certainty or maybe in fairness I don't think that uh Mike Leaven is certain of his interpretation of consciousness.

[00:13:39]It's simply that this null hypothesis is slightly different and in some sense the null hypothesis that thing that you believe to be true by default and you think that is more likely to be true and you need evidence to get you away from it. In some sense how do you get this null hypothesis? It's often a complicated conspiracy theory that you just happen to share with your environment. And uh for instance there are cultures in which uh people see uh are embarrassed that they see ghosts and like ours and uh >> uh there are other cultures where people embarrassed where they cannot see ghosts and in every culture there are some people who can see ghosts and people who cannot see ghosts.

[00:14:19]>> Right? This doesn't say whether ghosts are hallucination of psychotic people.

[00:14:23]But some cultures per default say ghosts are a thing and other cultures per default say ghosts are not a thing. And so what can move me over the evidential threshold in one way or the other is difficult. Right? It would mean that a spirit can exist without a body or without a single body and can exist across bodies maybe. Right? In some sense this is tool for the spirit of a state or the spirit of a city or a family that or a religion that can exist across many bodies. But can it be personalized and can it be conscious uh and agentic in the same way as our own mind can be that's that's a more difficult question and so as long as um I don't have a criterion to determine this I think I have to remain agnostic >> which means I have to admit I I really don't know this it's an open question to me both is possible I may think this is more likely for these and these reasons and this may be an experiment that would uh basically shift my belief and in this direction of having more confidence in the opposite understanding.

[00:15:28]>> I'm definitely agnostic as well. Of course, of course we I I can't know. We we don't know. But it it just it feels easier if you say that a frog is conscious. If you can't know that the clump of 2,000 frog cells is or isn't conscious, considering the frog is conscious or we're accepting under this set of preconditions that we think the frog is is conscious, it just it almost feels intellectually easier to just say, okay, well then probably the 2,00 cells are because then I just don't know where. So here's a slight argument that uh that yes so uh the reason why I think it might be possible that the body of is not conscious >> is that uh the body is limited in uh what it can do to itself. It seems to me that I can use my consciousness to restructure my mind to rewrite my own source code and uh sometimes it takes a very long time because there are so many dependencies and there's a lot of structure that I build in my childhood before I knew what I was doing and later I build structure that depends on me having built this child and now this child is doing things in me and is building the adult and then the adult and so on. But uh if I really direct my consciousness on all these things, I can go into the child and fix a few things and I can change the type of mind that I am.

[00:16:48]There are many different architectures that can exist on the subset of my brain and I can access a range of them using my conscious activity by to rebuild myself and I don't observe this in organisms and that if uh I lose an arm or a leg I cannot just regrow it. Some organisms can do this to some degree but these organisms tend to be more blobby and much smaller. And then uh I cannot also grow a third arm if I would find it practical or uh change my anatomy using the consciousness of my body even if that would be practical. So it seems to be that this process is more local adaptation where something is reacting to the direct pressures of a local environment and so this uh limited ability of my body to restructure itself rather than uh popping into a shape of a different phenotype. If the um environment demands it like I can do when I roleplay a different person or a different psyche or a different job right that that is something that I cannot do I cannot just change or roleplay my body into a different shape >> so this might be an indication that the mechanism that organism my body is not as powerful as the inflexible is the mechanism that is optimizing my mind so there's also no strong evidence to the contrary it's still possible that simply bodies maybe more rigid or harder to control or more stubborn. But uh it's also possible to say oh the body is organized according to less flexible less um self-aware less self-reflective principles. So the body might not actually know its shape as well as the mind is able to model the shape of the body.

[00:18:29]Yes, this still feels like more of a quantitative difference than a qualitative. Like I don't want to make the claim that the xeno the xenobot is the same amount conscious as the frog.

[00:18:40]You know it might be 1 one billion as >> it's local versus global information. So the xenobot itself does not actually seem to be structured in a way where the different parts need to know much about the other parts.

[00:18:55]uh and it's maybe possible for the >> maybe it's sufficient for the xenobot to know roughly in this direction there is a gradient to foot and then it tries to rudder itself there and then as an emergent phenomenon it's mostly moving in the same direction uh and uh maybe he has results that are showing that it's much more complex than that but for the frog the frog is able to discover a fly in the environment and target the fly and go after the And I suspect that a xenobot cannot make such a model of its environment and the place of the xenobot in the environment.

[00:19:32]>> And the kind of model that we can make of our environment is more complicated than the frog can do because we can do reasoning. We can probably represent more complex geometries reliably and plan our path through them and so on than the frog can.

[00:19:47]>> And this is the result of basically our conscious attention that allows us to make this global model of reality.

[00:19:53]Yes. I guess what I'm curious in understanding is what what you think the minimal set of biological ingredients is for consciousness. Like what's the absolute minimum ingredient list in terms of biological wetwware in terms of substrate that produces consciousness?

[00:20:10]And this is consciousness of any level or any degree. Not human consciousness, frog consciousness, but just the absolute minimal set of biological wetwware.

[00:20:21]I suspect it's uh multisellularity and incentives to form coherence and long enough evolutionary time spans in the niche in which these incentives can come into play. So I am very open to the possibility that plants or forests are conscious in the sense that it's what that they have a notion of what it's like and some notion of real time even if the real time is much much slower than ours much less resolved than ours but uh that is it's something where I think we can possibly find an answer via simulation experiments and uh for cells I'm less sure for a biological systems uh in nature. I'm also not sure, right?

[00:21:06]It's not completely out of the question that there are a biological systems that are conscious that imagine you have a gas planet and you have some storms on the gas planet and they uh exist for millions of years and they rotate metallic clouds and they create magnetic fields as a result to which some of the clouds react slightly differently than others depending on the metals in there and there is some feedback and this whole thing is driven by tidal dynamics and maybe at some point this wakes up and uh realizes, "Oh, I'm all alone in the universe." And maybe it's out of boredom. It creates the first cell is some kind of annoyment probe to colonize the universe. This cell is like this little robot. When you drop it on another planet, it's going to try to work and adapt itself and evolve and self-replicate. And in many environments, it might be able to take boot. So once you have the first cell, life seems to be somewhat easy. But uh to get to the first cell is very hard.

[00:22:00]Maybe there are a biological processes that are not self-replicating but that are still able to wake up but maybe uh blazera is correct and others and say that self-replication is the first thing and once you get to self-replication um sentience uh and intelligence follows right and that is the default hypothesis that most people in the life sciences currently have that basically self-replication is more elementary um in mutation than mindedness and agency and the other stuff basically results from this, right? But in in principle, logically, it's possible that it's the other way around and we just don't know at this point.

[00:22:43]>> The giant shoulder mission is to explore radical ideas in biology, neuroscience, and consciousness and elevate those stories to the highest possible level while keeping them accessible to everyone. If this interests you and you want to support independent science, then please consider subscribing to the clips channel. Check out our 26 neuroscience book.