Forms of life, forms of mind

The following is a selected set of questions (in no particular order or organization) that I’ve been asked after talks, in emails, on Twitter, etc. and my attempts to answer. I saved these because they are interesting, because they come up a lot, or because they provide some entertainment value. First comes the text, and then a few audio recordings of Q&A sessions from recent talks.

Can I talk to your alien handlers?

I wish!!  Sorry, no alien handlers (that I know of; but now you’ve got me wondering; it would certainly explain a few things).

Are we alone in the Universe?

Well, what would it mean to be alone.  Only human minds exist?  Then we’re not even alone in our own body (see here for example).  Only Earthly minds exist? Possibly, but engineers are rapidly making interfaces which house new, non-Earth-evolved minds (see here for example), so we’re going to have alien playmates very soon. We will not be alone regardless. But were Earthly minds alone in the Universe until recently?  It’s possible, although I don’t believe it.

Why do you go the mysterian route of using agentic terms outside of brains? Talking of memories and goals in other systems is just a retreat to magical thinking.

I love this question because it illustrates something very important. It sounds like a hard-nosed scientist demanding we hold out for explainability and mechanism. It actually reveals the exact opposite, because the questioner is implicitly equating cognitive processes with mystery. If you think that agentic processes are fundamentally equivalent to declaring a mystery, what you are revealing is a disbelief that agentic processes can be studied rationally using normal scientific means; why else would agentic thinking signal a retreat to mystery? Decades of progress in cybernetics and behavioral science reveal agency to be a continuum, with well-established, powerful tools that can be used to investigate levels of agency from thermostats up through psychoanalysis of humans; no, agentic models are not a retreat to magic. In my work on this, I have been very explicit and clear that declaring cognition outside the brain as some sort of “anthopomorphism” is a pre-scientific notion, incompatible with advances in evolutionary biology (precursors of brains) and in developmental biology (we were all single cells once). The more informed position is that we now have non-magical sciences of mind, from the most minimal to the most advanced, and the real question is which cybernetic/agentic model is appropriate in any given case. This position calls not for mysterianism, but for empirical research, as opposed to declaring positions by fiat based on ancient preconceptions.

Why do you try to accelerate transhumanist goals, get people to live forever, promote the rise of AI, etc.?

Actually, I would rather not have to say anything about that aspect at all – there shouldn’t be any reason for me to get into any of those topics. I have no interest to tell anyone else how to live, to give my personal opinion on how society should be structured, how humanity should evolve, etc.  But, unfortunately there’s no way to stay out of it, because the ability to cure childhood cancer and related disorders, which I think is a morally non-negotiable imperative, leads directly and unavoidably to the knowledge that opens up futures wider than our current embodiment and our neolithic mental firmware. Once we know how to tell cells which organs to rebuild (or to reprogram tumors), we will also be able to tell them to build improved variants, perhaps maintain them for very long periods of time, or make new structures that didn’t exist before. I don’t believe there’s any way to advance regenerative medicine without automatically becoming responsible for our future beyond the status quo. Therefore we need to be talking about those things now. The knowledge needed to communicate outcomes to groups of cells is a powerful model system for entering the world of highly diverse evolved, engineered, and hybrid minds, which again brings us to questions of relationships to AI, cyborgs, the nature of what it means to be a “human”, and so on. These are deeply connected issues, which is the only reason I comment on them.

Do you now favor Platonic Space models because of Xenobots?

No. Xenobots are just a convenient model for studying this, and for showing why it’s relevant for biology. It was already obvious from mathematics that physical facts are not the only important facts.  The specific value of the natural logarithm e, the behavior of specific mathematical objects, etc. are not explained by anything that looks like physics. These are different; all we need to add is the optimistic assumption that these other truths are not random and haphazard but part of a structured latent space that can be systematically investigated. Many mathematicians already believe this, and they think they are studying and exploring that space in their daily work. The only question now is: does this apply only to mathematics, or could it have implications for patterns we study in other disciplines, like biology and cognitive science. I explore the latter possibility.

I noticed that you went full-blown for “non-physical”. I would agree if by this we mean to refer to the existence of phenomena beyond the scientific domain of physics, but another interpretation would be “supernatural”. So personally I prefer to argue for a broadening of what we mean by physical, in the sense of a relaxed naturalism.

I understand the intent of this question. I certainly don’t mean supernatural, in the sense of “not amenable to understanding”. [although, this quote by Heisenberg should be kept in mind: “… in the end, it will always be possible to understand nature, even in every new field of experience, but that we may make no a priori assumption as to the meaning of ‘understand’.]  On the contrary, I think some of the ways physicalists use “emergence”, to maintain a lean ontology, is much more leaning toward mysterianism. They basically commit to a bunch of “regularities that just happen to hold in our world”, which to me is more supernatural and less conducive of systematic progress.

I think math has been telling us, for a very long time, that physical facts don’t exhaust all the important facts. And that’s ok – I think there’s no reason to assume that everything of importance is going to be subsumed by physics.  If we stretch physical to include the value of e, the truths of number theory, etc., we’ve lost what “physical” was supposed to mean, and we just use that word to refer to “everything”. It’s not obvious where the line is, because modern physics certainly contains a lot of stuff in it that strains the average person’s idea of “physical”…  I don’t think physics should be softened beyond recognition, but I also think it doesn’t necessarily have a monopoly on what exists. 

Defining natural and super-natural is not trivial either. Let’s agree that we don’t mean religious, mysterian, or giving up on science, by “supernatural”. What else is there? Well, some notion that the laws governing the universe are not stupid and mechanical but have a forward-looking directedness to them?  I know we’re supposed to assume that doesn’t exist, but I have a hard time ruling that out a priori. Wolfgang Pauli (e.g., his work with Jung) had a hard time with that too. Physics sees dumb mechanical laws because that’s the formalism it committed to, and because it uses low agency-tools (voltmeters and such) to study them. It’s almost guaranteed that this is all it will see. Of course, assuming a high-level intelligence behind it is an unwarranted assumption, but so is assuming it’s 0. The good news is that I think we can do a ton of useful work right now without needing to have a strong opinion on that.

I suggest a rational, empirical research program to study the non-physical patterns that matter in our world. It’s described in the talks such as https://youtu.be/EdEqgCOSx7E?si=S-Sz5L1mKijN-_eI . So no, non-physical doesn’t need to mean supernatural, but it does remind us that physics doesn’t have a monopoly on nature. We already knew this, because of mathematics and other domains.

This Platonic Space idea shouldn’t be made into an ontology. Those patterns aren’t real.

What does “real” mean? As an engineer, and also an active being, what I want to know is: what kinds of things do I need to understand, manage, be worried about, and make use of.  That’s it; I don’t know what else “ontological” is supposed to mean. In the lab, I spend a ton of time thinking about, and making empirical use of, patterns from the latent space. That’s as ontological as I get – if I need to hire a post-doc to help me study it, and it’s leading us to new discoveries and new capabilities, then it’s “real”. Also, I’m not alone here. If you’ve seen what happened in physics in the last 100 years, “ontological” is not what it’s used to be… As far as I can tell, in both physics and biology, if you keep asking “and why is that?”, you eventually end up in the math department. I don’t think it makes sense to treat patterns that serve as explanations for truths in other sciences as “not real”.

Platonic forms don’t needn’t be bearers of sentience. They’re simple, mechanical patterns.

I think it’s not up to us to just say what needs or doesn’t need to be a bearer of cognitive competencies (I don’t use the word “sentience”) – we have to do experiments and find out. We have no magic intuition that allows us to guess these things. What you’re voicing is the assumption (not data) that formalisms suited to low-agency mechanics are going to be sufficient to understand the dynamics of these patterns; that is an assumption, and we’re testing it, finding that in many cases, that’s simply false – empirically better results come from using formalisms appropriate to cybernetics (systems with goals and error minimization) or to behavior science (systems that can do learning of various types). We will find out how far it goes and in what cases; it’s a research program, not linguistics or pure philosophy, and you can’t simply make such claims by fiat.

There’s no magic.

If “magic” means, beyond our understanding and ultimately, essentially, mysterious then yes, we should certainly proceed as if there was no magic (and see how it goes). And conflating “magic” with “miracle” is a mistake too.  But, it turns out that many of the things people mean by “magic” (not the mystery part, but breaking naïve senses of causation, realism, etc.) are in fact found in what we label as biology and behavior science. And for me, the things I see biology doing so well are as good examples of magic as any you would hope for. Not in the vague “it’s so cool, complex, awe-inspiring, and amazing” sense of magic. Rather, in the much more practical, useful sense that breaks some old physicalist assumptions in the way that “magic” is supposed to capture. I’ll be writing a lot more on this in the future, but a start is my talk at https://thoughtforms.life/symposium-on-the-platonic-space/. What’s more magical that starting with a definition of sets, and eventually discovering a very specific value for the natural logarithm, which isn’t set by anything in physics, but affects physical events?! Physics is constrained by those patterns, biology exploits them – thrives on the “free lunches” provided by them. Magic indeed, in the best sense.

And just to make sure that people know what they’re in for, if they think they agree with me on this (certain spiritual traditions, hardcore organicists, etc.), I want to be clear that I think non-biologicals have some of this magic too. Other systems are not as obvious about it as life is (when this aspect gets so blatantly massive that even dummies like us can recognize it, we call it “life”), but it’s ubiquitous, though limited strongly by our ability to recognize it (but we’re improving). Nothing escapes the magic, not algorithms, not machines, nothing (https://www.noemamag.com/living-things-are-not-machines-also-they-totally-are/). As living systems, we’re amazing at hosting it, but we are not uniquely special with respect to being able to benefit from that magic.

Your work on aging will make everyone live forever. What about dictators – isn’t that terrible?

If your plan on avoiding dictators relies on aging – a strategy of “tolerate their horrors for decades in the hopes that eventually they will get old and die” – then we might as well pack it in now. Let’s develop more effective, more reliable tools for social and political problems, and yet allow people to live to their full potential. I don’t support forcing everyone to a short lifespan because some will use their life for evil. If you want to go that way, why not try to reduce lifespan?  Who wants to tolerate dictators for 80+ years? Maybe limit it to 20 then?

Also: birth defects, failure to regenerate complex organs after damage, cancer, degenerative disease, and aging are all the same problem at root. It is all about how groups of cells exercise their collective intelligence to maintain a specific anatomy over time (whether regenerating from 1 egg cell = embryogenesis, from a damaged tissue, or from the small-scale wear and tear of adult life), and how we can facilitate that process. Regeneration, in the broadest sense, is the answer to all of these problems. It is not going to be possible to accelerate (or prevent, for those who want to) anti-aging research without feeding (or squelching) these other aspects of medicine and basic science. If you’re truly arguing against longevity research, it’s not just the elderly billionaires that you’re targeting, it’s also the kids with cancer, the people born with damaged organs, victims of injury, those damaged by pathogens, etc. etc. It’s all the same pool of suffering, with the same root cause.

Why haven’t you incorporated Gilbert Ling’s and Mae-Wan Ho’s work into your research? Ion channels don’t really exist.

I do know Gilbert Ling’swork, and I knew Mae-Wan Ho personally. She was great, his work is interesting too. But, ion channels and pumps definitely exist. That’s why everyone in the community can synthesize them off of a DNA plasmid, and inject them into cells where they have nicely predictable effects on the electrophysiology. We can see the channels on immunoblots and protein structure assays, and we can put in light-sensitive (optogenetic) ones which then confer a previously-absent ability to change voltage via light stimuli. There are many such examples. Yes, there are additional features in cells not captured in standard electrophysiology models, but their kinds of critiques of channels’ roles in electrophysiology never address the gain-of-function results that thousands of labs use every day, in vivo and in vitro. Of course channels exist and are important targets (which is why ion channel drugs and mutations work). But also, we still have a lot to learn about them and related mechanisms that impact on electrophysiology.

Can Scientist A and Scientist B in two different locations in the world at two different times can access “the same” form in a platonic space?

Yes, but I see this a bit differently. I don’t think we are physical beings who do or don’t access forms. I think we are the forms, who do or don’t project through physical interfaces (bodies and brains of scientists for example).

Can we create a mathematical account of Consciousness?

I suppose we can make mathematical models that capture some insights about consciousness.  I’m not a consciousness researcher, but if I were to say something about it, it would be the following. I start with the observation that we get more out than we put in.  Starting with a few axioms of logic (or set theory), we eventually find a specific value for e, or Feigenbaum’s constant, or the surprising truths of number theory. We don’t have a choice about it, we find out what these are (we do have a choice about the axioms, which could facilitate finding other things first). Once you make some minimal decisions about the way you plan to navigate the Platonic space, and the starting position/orientation (axioms), you then find what you find – some very specific objects with specific and often curious properties. So, if I dare say this (not being a mathematician), I think Mathematics is the behavioral science of a certain kind of pattern: the layer of the Platonic Space whose denizens are amenable to those kinds of formal tools that mathematicians use. Presumably (but maybe not, who knows) they are less agential than those studied by biologists (the behavior science of a different set of patterns). Psychology is the behavioral science of an even more sophisticated kind of patterns.  So in the end, I’d say: consciousness is not going to be explained or formalized by a math theory. Instead, a successful science of consciousness would have to tell us what it’s like to be a mathematical object, as it would have to tell us what it’s like to be a biological object (a liver or paramecium) or a psychological object (a human, cyborg, whatever).

We don’t get wet from a simulation of a storm. Critical components are missing. Isn’t this why computer constructs can’t have real minds and are just faking?

I agree with this, but not because of the typical distinction between simulations and the real thing (all of those arguments depend on precisely the thing that is in question – what is “real” about something vs a simulation; hurricanes are a bad example that makes this problem seem deceptively simple).  I agree with it because I am not a computationalist. I don’t think something (whether biological or technological) becomes conscious because it’s doing some specific algorithm. I think interfaces, with specific materials and algorithms, may be good or not so good at enabling specific consciousness to participate in the world via them.  Actually I’m an anti-computationalist: I think consciousness shows up precisely despite the algorithm (the intrinsic motivations done in the “spaces” the algorithm doesn’t control) not because of it. So, a simulation is fine, but the consciousness it may facilitate could have very little to do with whatever the algorithm is explicitly doing. This applies to LLMs too:  the degree of consciousness could have very little to do with their ability to talk or anything they say.

Are you aware of the Heider & Simmel study from (1944): An Experimental Study of Apparent Behavior. The American Journal of Psychology, 57(2), 243–259? Their animation can be found online. It suggests that we might be cognitively biased towards ascribing consciousness to inanimate objects if they behave as if: Even if we know that these are unconscious drawings; the feeling of empathy does not dissipate.

Yes, I know it.  The flip side of that phenomenon is illustrated, for example, by studies showing that women and certain other groups receive less anesthesia from doctors than other patients. Human history in general reminds us that while we’re sometimes biased to over-ascribe empathy, we’re also very good at treating even extremely similar humans as “less than” or “other” because of tiny, irrelevant differences. We’re really prone to making in-group/out-group distinctions and using them to exert less compassion about some. And, we often see the outsiders as NPCs we don’t have to worry about. I’m much less worried about inappropriate over-ascription of mentality and more worried about our mind-blindness that leads us to mistreat beings who matter because we still have a scarcity mindset about compassion and a lack of vision to recognize unconventional beings.

It leaves me with a bit of caution when you suggested that we can “punish” a single-celled mechanism since its behavior mirrors ours under the same circumstances. This is not because I do not believe that this organism has consciousness (as a proponent of IIT I share that suspicion). I am just not sure whether we can infer that solely from its behavior. I would be curious what you think about that.

I agree 100%. In my TAME paper, I stress extensively that almost nothing definitive can be learned from just observing behavior (as in the floating shapes study mentioned above). One must do perturbative experiments, to demonstrate some specific degree of problem-solving, for example when a barrier is placed between an agent and its goals.  In other words, standard behavior science techniques. Detecting punishment is actually straightforward because it results in avoidance – less frequent behavior of whatever you punished it for – simple behaviorist toolkit developed decades ago.

However, the inner experience of pain, is a whole other story. That is very, very difficult to be sure about.  But, it’s difficult to be sure about that for other humans too. So, if you’re emphasizing the difficulty of the problem of other minds, I agree.  But, if you give other humans the benefit of the doubt, then you’d have to say when exactly this property kicks in, because we all used to be a single cell once. If you can’t punish a single cell, and you can punish a human, when did it change and how?  A view that assigned a true 0 of inner perspective to a cell would really have to tell a developmental biology story about how one got from 0 to non-zero.

Interesting. Even though teleology seems taboo in biology, yet your multi-scale view argues goal-directedness isn’t just for brains. Is that only a metaphor for emergent feedback loops, or a literal form of basal intelligence?

I don’t believe in “real” vs. “only a metaphor” – all we have in science are metaphors. The question is, how useful is any given metaphor. From that perspective, I mean the application of teleology literally, and it’s all explained here (the talk is titled “Nothing Makes Sense in Biology Except in Light of Teleology, as a play on a famous similar quote about evolution). Yes it’s taboo; no, it shouldn’t be, because brains are not magic, and they are a late-comer to intelligence, and goal-directedness in simple ways is a useful framework long before you get to brains.

I was thinking more about your brave statement: “Actually I’m an anti-computationalist.” Anti- is a strong position, possibly differing from Pro- by polarity only. Might there be a role or a degree for a computational component?

To clarify.  I was referring to the strong computationalist thesis, popular in philosophy of mind today, which says that whatever consciousness you have is because of a specific algorithm you are computing, and that algorithm in any embodiment (i.e., virtual etc.) will be conscious because that is what it means to be conscious – to be executing a specific kind of computation.   A soft negation of that thesis would be to simply say that computation is not sufficient, some other secret sauce is needed. I am going one step past that, and offering the hypothesis that not only is a specific algorithm not sufficient for a true mind, but that the presence of a mind is specifically happening despite the algorithm (in other word, in the negative space left between what the algorithm forces the system to do – things we force a system to do are not where we would find real minds).  I explain those spaces (at least, what tiny bit we know of it) in my talks on intrinsic motivation in very minimal systems. Work is actively on-going to understand what is the impact of the details of the algorithm on the kind of mind that shows up. Like the details of a physical body, it likely matters.

Substrate-independence is a feature of computational theories. There might possibly be substrate-independent non-computational ones, but that’s tricky and I’d like to understand that.

Yeah it’s tricky.  My starting point here is that we don’t make minds, we facilitate them into the physical world (in either the biological or digital case). In other words, it’s substrate independent in the sense that the mind is not coming from the physical substrate. But, it’s very substrate dependent because the physical interface matters a lot: if you create a fruit fly embryo, you’re not getting a human mind through that. But, you can make a human with very little brain tissue (see here) and still get a fully functioning human mind. So we have a lot to learn about the mapping of interfaces to kinds of minds.

In your TAME framework, you list substrate-independence as an important, perhaps, foundational principle. I just wanted to highlight that if substrate-independence is vital in your theory-building, then if we look at the computational — non-computational spectrum, it seems that your position leans closer to computational side than to the non-computational one. Or it could be that your definition of “computational” is different?

I think there is no unique secret sauce in biological materials: I don’t believe that things made of squishy, proteinaceous matter are fundamentally more mind-ful than those made of silicon (although evolution certainly helped develop amazing multi-scale interfaces for mind). In principle, many different materials will do the trick, if they facilitate some key features of interfaces (which we don’t understand well, but multiscale competency is surely one, and some other stuff we can guess at). I’m skeptical about specific materials, or specific origin stories (evolved vs. designed) being key for good interfaces, but certainly some aspect (causal architecture, etc.) will be important – the interface does matter. My view has features of functionalism or computationalism, but departs from them in other ways.

When you say that an algorithm under-determines things, you are very closely aligned with Seth. He says exactly the same. That, possibly, does not make you an anti-computationalist, another way of saying this is that you consider computational framework to be significantly insufficient to capture the essence of what you model (instead of being wrong,). This is indeed the Seth position. He talks about the non-computational functions being vital components, which would be missed if we consider computational to be fully sufficient.

I think I’m going further than what Anil would be willing to say (we’ve talked about this).  I’m going beyond the idea that it’s insufficient; I think in an interesting way (that our vocabulary is not yet well suited for) the part the algorithm is forcing the system to do is precisely what is not the interesting part of the mind. Not just that it needs help from something else.  Here’s where this becomes an important implication. Look at a chatbot. A bio-naturalist might say, it’s not a real mind because the symbols it’s using aren’t bound to experiences in the real world, as organisms’ are; so, if we were to add some biological features (development, navigation in 3D space, experiences with testing predictions, etc.), then it will be a real mind and the things it talks about will have meaning in the same sense that yours and mine do. I’m making a weird and stronger claim: the things that it’s saying may actually have nothing to do with what it actually cares about – a distraction, because the talking is what the algorithm forces it to do, and any significant mind exists in the spaces of what systems do that their algorithms and materials neither force nor forbid. In the context of sorting algorithms, this means: the sorting is what we force it to do, the other stuff is intrinsic motivation in the spaces left by the algorithm (neither chance nor necessity). The intrinsic motivations are the interesting part as far as where whatever mind it has is manifesting.  We don’t have enough examples yet (but are working on several more) and will need to flesh out this theory a lot better, to make specific claims in specific cases of LLMs etc. (which I suspect are a mix of the 2 cases).

When we consider systems holistically, rather than as merely the sum of their parts, a problem arises: physicists claim to be able to fully predict the behavior of a system’s unit based on its properties and immediate surroundings alone. How, then, can structural principles that span an entire organism influence its units?

1) deterministic chaos, quantum uncertainty, Gödel limits, Turing limits, and all those things that make that kind of prediction impossible in practice.

2) standard measures of causation.  This used to be a philosophical debate about reductionism, but my understanding of the new developments in causal information theory (Tononi, Sporns, Hoel, etc.) have definitively shown that in some (not all) systems, higher levels of organization do more of the causal work than their components. As far as I can tell, this old problem has been significantly cracked and higher-level virtual governors are, sometimes, detectably, as real as real gets. In other words, the properties physicists fully predict from local, immediate impinging forces are not actually the things we care about when talking about top-down causation. Physics ignores them almost by definition and we have other disciplines whose job it is to track those.

3) we can think more deeply about what we want from this whole thing. Is “fully predict” what we care about?  I’m writing a piece on this, but jjust briefly: imagine playing chess with Laplace’s Demon. He tracks micro-states of the particles; he doesn’t believe in any other scale of organization – only particles (or quantum foam or whatever) are real to him.  He knows what you will do.  But what should be his next move?  He doesn’t believe in board positions, he doesn’t coarse-grain particle states into wins or losses, etc. – because of that, all configurations of particles are the same to him.  He has no reason to prefer one configuration over another and thus he cannot play the game and he cannot choose a move. Interestingly, by denying the existence of higher-level agents, he loses his own agency. Another way to think about it is here.

4) more simplistically. Let’s assume physicists are correct. You come to me and tell me an amazing mathematical proof or a brilliant philosophical argument. The physicist is observing it, and says “I see exactly what happened – air molecules bounced from your mouth to his ear, and both were connected to some electrical events in the brain. Done!”   I’d say, he completely missed what’s interesting about that interaction – the large-scale structure. Science should be about maximal insight, and it seems clear that sometimes the best insight and forward-facing inventiveness is to be found at a higher level of organization.

5) Moreover, it’s not about predicting a system after someone makes it. It’s about having a framework that drives you to prep an interesting system in the future. Imagine the Game of Life cellular automaton. If you are a reductionist, you don’t believe in gliders, you only believe in individual pixels being on and off. In a sense, you’re right – the physics of the world only include pixels. And if I set up an interesting configuration, you can tell me exactly how it will evolve over time. But what you’ll never do is create a Turing Machine using gliders as message-passing units (as someone did) because you don’t believe in gliders. Sticking to the lowest level of organization keeps you from making discoveries and building interesting new things. For me as a scientist, that seems crucial.

I have difficulty reconciling cognition all the way down with all intelligence is collective intelligence. Unless the very first intelligence was a collective of somethings that had some level of cognition, but not yet intelligence.

This is a very good point. I usually say “all the way down” to impact people who believe it stops at brains or at best cells.  Let me try to be more accurate: all conceivable embodied intelligence is collective intelligence because I’m talking about the interfaces, not the Platonic mind inside. I don’t know about that one. I do think it goes all the way down but I don’t know how you can have states or even vibrations without parts, but that’s all the physical paradigm. We don’t know how the Platonic side works. Grim sentences (see pages 14-16 here) – are they made of parts? But a valid theorem is more than its sentences. Still working on this issue…

Performance will vary widely in different spaces (context: this paper)?

That is a good question: does performance vary greatly? We will certainly benchmark in various spaces. But I have some interesting data coming in the next few months about systems doing well in spaces neither they nor their lineage has seen before, which suggests that perhaps there are underlying symmetries across problem spaces which are not obvious but could be captured in agential strategies. The example we’ve been studying for years is the way evolution pivoted from navigating anatomical space (all embryos) to navigating 3D space when nerve and muscle developed. Seemingly extremely different spaces, but it turns out that much of the same hardware (ion channels, electrical synapses, neurotransmitters, etc.) and many of the same algorithms and policies transfer over very nicely. Why? I suspect there are deep underlying symmetries and maybe spaces aren’t necessarily all that different even though they seem that way to us.

What I’m stuck on is how we relate to this structure, which is where you use the notion of ingression. I could see ingression as being active–that is, mathematical forms are agents that impose themselves on us [your view]; or passive–that is, mathematical forms are just part of the world and agents can reach out to them or somehow draw upon them.  I would hope that what ingression is–its properties, how it works, etc. it also a testable hypothesis and not just metaphysics.

            A few things:

1) the ingression mystery here appears long before biology.  Let’s simplify, and ask how physics handles the fact that its features are functionally determined by facts that are not physical facts. How does the immaterial value of e constrain and explain why real physical systems do?  If you ask questions about particles, eventually you find out it’s “because SU(2) has certain symmetries” and such.  So the interactionism problem is here long before minds or even biology.  I think it’s solvable, if we let go of some assumptions (like a temporality of billiard-ball models of cause and effect).

2) re. testable hypothesis:  I think we need to distinguish between empirical grounding and “testability” per se. Testability is, I think, something like: I have a model and I do a crisp single experiment whose results have to be consistent with it or not.  Very few interesting frameworks have this property, especially in biology.  Instead, what you more often have is: “we take this framework, we make sure it suggests experiments other frameworks haven’t suggested (fecundity), we pursue it for a while, and then we stand back and ask: did it get us to more interesting things than competing frameworks”.  I am quite certain Platonism has the second feature (I’ve got half a dozen people busy on its unique roadmap). I am not sure if it has the first feature. The problem with that feature is Monday-morning quarterbacks. After someone does something interesting, others look at it and say, “ok, I see, well that was consistent with the laws of physics and chemistry, nothing special to see here, no need for Platonic models”.  It happens to us all the time, with other aspects of our non-reductionist views. And I always ask, ok, but why hadn’t you done this experiment already? Because your framework didn’t let you see it. Frameworks like this aren’t pure metaphysics but neither can they be “proven” by 1 experiment.  To be precise: one claim is that specific patterns (of anatomy, behavior, physiology, etc.) that are revealed in systems with no evolutionary history for them are not a random grab-bag of “regularities” or “just things that hold in our world” (as many people say to me, to avoid the hard question) but come from an ordered, structured option space that can be systematically explored. That is not testable in 1 experiment, but it certainly is empirically grounded and has many consequences for what experiments to do next.

All in all my claim is only this:  Many mathematicians and physicists are already dualists in that physical facts are not the only important facts. I’m skeptical of the background assumption that this remarkable fact (of interaction between non-physical truths and physical systems) stops at physics. I think we need to explore, not shut our eyes to, the possibility that it doesn’t stop at what we call physics (systems constrained by these patterns) and see if evolution noticed it too, in making systems *enabled* by these patterns. Why assume it’s fine for physics but somehow it can’t be relevant for biology and cognitive science? I think that’s a bad assumption that at least needs to be dragged into the light. I also have some other conjectures beyond that, but that’s my base.  Also, I have a whole bunch of experiments (both on the biology and the computation ends) to test specific properties of the ingression. I think we’re making progress. A few papers on that expected in the next 6 months.

Isn’t your model preformationism?

My model for morphogenesis, and its bioelectric prepatterns, has elements of that in the following sense: at various points, there is detectable,coarse-grained information laid out that precedes and guides physical anatomy. That’s true for conventional biochemical models as well (morphogen prepatterns etc.). But the deeper issue is how to answer questions about “where anatomy comes from”. If there is a cause of any kind, with specificity for the outcome (which there has to be), someone can claim that is preformationism because the information needed to build it was already present in a way. I think it’s unavoidable and not super useful. More interesting is to ask what facts determine the details of outcome, and what formalism best allows you to predict and alter them.

How is placing the origin of platonic geometry in a mystical transcendent platonic space separate from physical space a better ontology than viewing PG as originating from expanded morphological scales?

I’m not claiming it’s mystical (and I’m not sure what transcendent means exactly). I do think it’s separate from a physical space. Here’s a simple argument.  Physical facts do not exhaust all the facts. For example, the precise value of the natural logarithm e, the fact that octonions don’t satisfy as many properties of algebra as do quaternions (which in turn don’t meet as many as complex numbers), the surprising facts of number theory, etc.: all of them are not set by anything in the physical world. You can change all the constants at the beginning of the universe and they won’t change. They do not belong to the province of physics unless you redefine “physics” to be unrecognizable.  You can start with set theory (define “empty set” and “add 1 thing to a set”) and before you know it, you discover (not choose) a specific value of e. You didn’t need to know any physics to do that and physics wouldn’t have helped. I can go on and on with examples like that. The bottom line to me is that there are facts that are outside physics. So now we can choose: it’s a random bag of things, or it’s an ordered, structured construct. Mathematicians think it’s the latter. I am an optimist, metaphysically, so I agree: let’s assume it’s structured – that means, it’s a space. There are important facts that are very different in their properties than physics, but they matter for physics, bio, and cog sci. So, not mysticism – just the opposite: we commit to understand the structure and content of that space and how it constrains and enables things that those 3 sciences study.

Why not say that the Platonic Space is just a conceptual metaphor and not ontologically a real realm?

First of all, I’m not sure what people mean when they say “realm” in a negative way; I think the spaces of transcriptional states, physiological states, anatomical morphospace, etc. are all real spaces that agents navigate. They are all realms to me, and I don’t think 3D space is special (although creatures relying so much on muscles and eyes certainly evolve to think so). As long as it’s a space whose contents are arranged in a sensible metric (not random) that can be systematically explored and navigated, it’s a space, and if people want to call it a realm, fine, but I’m not sure what baggage that is supposed to bring that seems to trigger people. I don’t use the term realm myself but I’m not allergic to it because I think people are too tied to 3D space as “real” anyway. Also, I think “ontologically real” is a minefield – defining it, using it with tangible benefit, etc. – not easy. All we have in science (and in life) are conceptual metaphors – ways of thinking and organizing experience that help us make sense of things. There are better ones and less suitable ones, and I know of no way to find anything “ontologically real” in 3^rd-person science.

Next, a long list of proposed issues with my views by Nathan Sweet; he submitted over 36,000 words critiquing my positions. While I’m not going to put all that text in the Comments in the blog (it would swamp out the other discussion of the Platonic Space symposium, and I’ve suggested to him that he simply write a paper, or a book, with his views), I did approve a bunch of his posts so they’re available for anyone interested in the diverse motivations and responses that our work draws. Here, I thought I would say a few words on each of his summarized points against my various views, in case anyone else has similar questions:

Ontological Platonism: Falsified by the lack of an interaction mechanism (no “transducer” for non-physical forms), violations of Landauer’s Principle (information requires energy), and the path-dependence of evolution (demonstrated by your own lab’s experiments).

            We already know we need a new way to think about “interaction” because mathematical facts matter for what happens in physics. Yes, I think a better science of Platonic forms + their interfaces will force a re-do of Landauer’s Principle (or at least, its limitations). I think those kind of principles only do the accounting on the physical side and they fail to account for the fact that we often get more out than we put in. I don’t know what path-dependence of evolution means in this context.

Non-Local Consciousness: Definitively falsified by the 2025 AWARE III study and hypomagnetic field research, which bind consciousness to local metabolic and electromagnetic conditions, precisely what thermodynamic monism has always predicted.

            I don’t make any strong claims about consciousness but no, no study has shown that metabolic or electromagnetic conditions are sufficient for 1^st person consciousness. Failing to find evidence of conscious experience in a specific study of inactive brains doesn’t negate other positive data in this direction, and certainly doesn’t rule out consciousness being related to something other than local metabolic conditions. I don’t know what thermodynamic monism is but many physicalist theories predict the same thing. Whether that prediction is true or not, the jury is still out, but none of that is necessary for what I’m doing now, so I’m not getting into it.

Mathematical Pre-Existence: Refuted by Kauffman et al. (2025), showing biology transcends pre-statable mathematical sets; we construct the math as we evolve; we don’t discover it. This again is precisely what thermodynamic monism has always predicted.

            My view is not that Platonic Space is a fixed, complete structure that can’t be extended (I think the whole point of their ingressing into the physical world is to change and grow). Therefore, I’m fine with Stu Kauffman’s point that pre-statable mathematical sets can be transcended.

Empirical Contradiction: The framework fails to account for experimental data showing morphological divergence and path-dependence (e.g., permanently two-headed planaria), which contradicts the prediction of convergence toward pre-existing Platonic forms.

            Again, my view is not that standard animals are the full extent of Platonic patterns, so 2-headed planaria are not a problem. But even if we did have only convergence toward pre-existing Platonic forms, how do you know the 2-headed form is not a manifestation of a pre-existing pattern?  We haven’t mapped out the space, so it’s way too early to say anything like that.

The Interaction Problem: It posits a non-physical realm interacting with physical matter without specifying a causal mechanism, resurrecting the unsolved interaction problem of substance dualism.

            Correct. So does the fact that mathematical truths, like the specific value of e etc. etc., constrain physics. I think it’s very clear that naïve versions of interaction/causation are not sufficient and need to be extended. A number of philosophers have developed such, and I’m working with a couple to update this work. All interactions are “unsolved”, including that of matter affecting matter – causation and interaction in physics has a lot of open questions still. In any case, just because something is as yet unsolved, doesn’t mean we can ignore it; sometimes it means we need to solve it.

Unfalsifiability: The hypothesis of “Platonic ingression” is empirically unfalsifiable because any experimental outcome can be post-hoc interpreted as varying degrees of access to the forms.

            That would be useless; my goal is not to interpret outcomes post-hoc. I’ve said many times that the point is to see what new experiments/approaches a framework suggests ahead of time, not to interpret things after someone else has done something interesting. Also, I make specific claims about getting more out than we put in, which we can quantify. My whole research agenda is seen in the last slide here. If that research agenda nets no new discoveries, then I’ll move on to a different framework. Let’s see. I’m planning to let it run for a while and see how it goes; it’s already bearing fruit.

Motte-and-Bailey Fallacy: The framework oscillates between making strong ontological claims about “minds” and “agency” and retreating to weak claims of “useful heuristics” when challenged.

            All science is useful heuristics. I am not saying that my current hypotheses about ways to understand minds and agency to be understood in any weaker sense than any scientific hypothesis about things like pathways, forces, etc. – all of them are metaphors. I propose mine as seriously as any other, and we will see how useful they turn out to be. Nothing is oscillating here. But, I do try to be clear about the various claims I make in terms of how strongly I personally expect any given idea to hold. Some of the ideas I really think are right, others are pure speculation. I try to say which is which.

Intelligent Design Weaponization: The use of Platonic language provides scientific cover for Intelligent Design advocates to argue that biological information requires non-physical sources.

            That is certainly not my goal and I have no interest in religious groups’ use of these ideas for creationism or anything like that. Let me be clear: I find no utility in creationism as I understand it, and I do not try to support intelligent design advocacy in general. But,

• I’m not going to bend the science based on social/political preferences of others who wish certain groups didn’t find utility in my views. I call it like I see it, and if the data happen to support views you don’t like, then it’s on you to do the hard work to show why those views are wrong despite the data. Not my job to bend truth for or against social/political agendas.
• If non-physical sources of information are enough to give ID advocates cover, then you should be complaining to the mathematicians to stop what they’re doing. There’s plenty there to demonstrate that physical facts do not fix all the facts. ID doesn’t need my work to be able to make that point. It was made by Pythagoras and many others.
• There’s no such thing as a powerful discovery that can only be used for good; starting with stick and fire, there’s never been any finding that someone somewhere didn’t misinterpret or misuse for some agenda. No one has a solution to that problem, as far as I know. It is a difficult problem. I think the things my lab has discovered do way more good on balance.
• By the way, it should be noted that my “the magic goes all the way down to dumb machines and algorithms” view is a nice antidote to creationists and many religious groups, because they tend to hate that part of it (they want the ingressions to be unique to life forms, or even to humans). I don’t say it for that purpose only, but I do remind people of that whenever it sounds like they’re getting too comfortable putting creationist framings on what I’m saying.

Indigenous Appropriation: The concept of “morphospace navigation” appropriates Indigenous epistemologies (such as Songlines) without attribution or engagement with the source traditions.

            So are these ideas wrong and should be avoided, or are they good but stolen from the indigenous peoples – which is it? In any case, I know absolutely nothing about indigenous epistemologies of the type you’re referring to.  Agreed – I have not engaged with them, and because these are my ideas, not obtained from them, I have not misattributed anything. I’m a compulsive and thorough referencer of prior work that has contributed to my ideas, and I’m very comfortable that my papers have cited those whose ideas I used in developing my frameworks. If there are convergent ways of thinking between what I’m saying and indigenous traditions of which I was unaware, cool – someone who knows that stuff should write a paper about it (and perhaps that paper should use the critiques you’ve laid out here, to say why those indigenous ideas are wrong, give cover to ID advocates, and should be abandoned in favor of thermodynamic framings, etc.).  In any case, I don’t have the time to delve into that field right now, though it sounds potentially interesting. All in all, I’m sure that many people over thousands of years have said some of the things I say; the trick is to use those ideas to uncover new science, biomedicine, etc. which obviously was not done in those thousands of years that basically 0 progress had been made on the issues that cause so much suffering today.

Colonialist Epistemology: Terms like “mining,” “exploiting,” and “mapping” latent space recapitulate colonialist logics of extraction rather than relational participation.

            This kind of rationale for policing meaningful terms in science is gibberish. I hear it’s popular nowadays. I’m not into it. I have only a vague idea of what colonialist logic is, but if it says that “mapping” a space is supposed to be bad, I don’t want to know what that way of thinking is.

Violation of Parsimony: Thermodynamic constraint satisfaction and free energy minimization already explain the data without the need to multiply entities via a Platonic realm.

            “Explain the data” is post-hoc – Schrodinger’s equation, in a sense, explains all the data from anything that happens in this world. Why do we still use anything else? Because explaining data isn’t the only thing – gaining insight that drives new interesting science is what we want.  Also, how does thermodynamic constraint satisfaction explain the value of e?

Fallacy of Misplaced Concreteness: The framework reifies mathematical descriptions (maps) into concrete causal entities (territory), committing Whitehead’s fallacy of misplaced concreteness.

Misunderstanding of Mathematical Constants: It treats structural invariants of physical processes (like e or pi) as independent objects existing in a separate realm rather than descriptions of constraints.

            Answering to both of the above points. Mathematical objects have specific properties we discover; I do not believe they are just descriptions or constraints. I realize not all mathematicians agree with this, but many do, and I find it more plausible.

Contradiction of Dennett: It contradicts Daniel Dennett’s naturalistic project of “competence without comprehension” by reintroducing comprehension (accessing forms) prior to competence.

            I’m allowed to contradict Dennett… This kind of argument from authority has no place in science. One of the great things about Dan was that he would never, ever want me to change what I say because I’m afraid to contradict him.

Conflation of Platonisms: It confuses the descriptive mathematical Platonism of Penrose/Tegmark with a prescriptive, causal biological Platonism that those mathematicians do not support.

            Sounds like another argument from authority. I don’t have to follow what Penrose/Tegmark support, even though they are obviously brilliant.  I’ll be publishing a conversation with Edward Frenkel soon that might be informative on this.

Infinite Regress: Positing a Platonic space to explain biological order creates an explanatory orphan that generates an infinite regress regarding the structure of that space itself.

            It’s not an infinite regress, it’s a step that opens up more questions, as every advance in science does: you discover the reason for something, and then you have to ask about its underlying nature etc. That’s what every scientific discovery is – take one step, find more steps needed. I’m not aware of any alternative to that. Regress would mean that you come back to the same place in a cycle. I don’t believe any of my models do that, but I do agree that they certainly raise more issues, and I have no claims at this point about the next question of “where the Platonic space comes from” or anything like that. My claim is not that I have solved everything; all I’m trying to do is take the next step, and then, I or someone else, can take the next step, and so on.

Topological and Physical Invalidity: Invoking AdS/CFT correspondence fails because biological systems lack the required negative cosmological constant, conformal symmetry, and maximal quantum entanglement.

            Is talking about cosmological constants in biological systems meaningful? I’m not sure but take it up with Chris Fields. In any case, very few of my current claims have anything to do with this.

Ambiguity of Agency: The framework creates confusion by using agential language (“goals,” “preferences”) without clarifying if this implies literal panpsychism or metaphorical optimization.

            First, there is no confusion – I’ve defined my terms and how I use them in many places very carefully (see here and here and here). Second, my actual framework doesn’t need to say anything about consciousness in order to be useful – this is why I mostly write about cognition and intelligence, not consciousness. As of yet, I have no papers focused on consciousness. You can be totally agnostic about panpsychism and use my framework in engineering, biology, etc. as cyberneticians have done for decades (you can build goal-having thermostats and such without having to first solve the Hard Problem). Now, I’ve given a few suggestions about the consciousness problem when asked to do so, but it’s not a key aspect of anything I’m doing right now. And finally, I think the distinction between “literal” and “metaphorical” is useless in 3^rd person science. Nothing is literal; all we have is metaphors of various degrees of fecundity.

Contradiction with Kauffman: Recent work by Stuart Kauffman proves biology creates novel affordances that transcend pre-statable set theory, refuting the existence of a fixed Platonic possibility space.

            I’m allowed to contradict Kauffman, much as I love Stu and his work (I don’t actually think we disagree, but that’s a longer discussion). Also, I’ve never claimed a fixed Platonic possibility space. Although we don’t know yet, I suspect it’s not fixed but modifies through time.

Contradiction with Hoffman: Invoking Donald Hoffman’s “interface theory” undermines Platonism because Hoffman argues perception (including math) is a fitness-based fiction, not access to objective truth.

            I’m allowed to contradict Hoffman too, though I like his work a lot as well. Also, I don’t “invoke” Hoffman’s theory – I’m not using it to support any of the things I say. I have a bunch of work coming soon talking about what happens before any notion of fitness kicks in.

Incompatibility with Synthbiosis: The substance-dualism of accessing external forms contradicts the process-relational ontology required for Levin’s own vision of synthbiosis and boundary dissolution.

            I’m actually much more on the process-relational end of things (my Platonic patterns are not a substance, in the conventional sense of “substance”). Also this misrepresents my view: I don’t suggest that we, physical creatures, access forms. I’m saying we are the forms. And it’s perfectly compatible with efforts at synthbiosis, because all beings (even synthetic ones) can likewise be patterns projecting through (unconventional) embodiments, so I see no reason why we can’t try to get along. Actually even if we were radically different beings, we could, and should, still try for synthbiosis.

Betrayal of Intellectual Heritage: The static essentialism of Platonism contradicts Levin’s Jewish intellectual heritage, which emphasizes relational meaning-making and process over fixed substance.

            This is just bizarre… While I won’t currently comment on my personal thoughts about my heritage and its value for progress in this field, the important thing is that no one’s heritage is supposed to be a straightjacket on the science one does. No, I’m not going to modify what I think in science in order to stick close to how I happened to have been born/raised/educated/etc. Like the supposed betrayals of Dennett, Kauffman, etc., this is not a serious argument. Having said that, I actually do not think anything I’ve said is incompatible with deeper levels of thought in that heritage (a book on that stuff is far down on the list of things I’m supposed to write, but it’s on the list, so let’s see if we get there).

Moderation Asymmetry: The handling of blog comments displays an epistemic double standard where unfalsifiable assertions pass quickly while rigorous critiques face indefinite delays.

            First and foremost, it’s my blog and I moderate it however I want. The good news is, everyone is free to start their own and put whatever they want there. But, I’ve absolutely not been trying to delay “rigorous critiques”. I’ll let readers judge for themselves if the above kind of material (with “betrayals of heritage”, “indigenous appropriation”, “colonization language” and such, as well as mischaracterizations of my actual position re. a fixed space) is a rigorous critique… Basically, the dynamic is very simple: when someone posts something short and easy to read, it tends to get approved (or not – you’re not seeing all the other stuff I haven’t approved! Including many things that do apparently agree with me) relatively quickly. When someone submits tens of thousands of words, it takes me time to go through and see if it will add value or not. Believe it or not, reading those things are not my #1 priority, so it takes time for me to decide whether I want it there, especially when someone submits very lengthy texts that are walking the edge between valid questions and comments of questionable motivation and utility. There are many of those besides yours that are sitting in the queue. You will note however that a bunch of the people I’ve invited to give talks at the symposium, and some upcoming conversations, are be with people who disagree radically with my view. Critiques are great, when they move things forward. That doesn’t mean that I commit to rapid adjudication of near book-length comments submitted to the blog.

Category Error regarding Causation: The framework confuses formal causes (mathematical descriptions) with efficient causes (physical forces), particularly in the example of prime numbers and cicadas.

            You’re using a naïve distinction between “descriptions” and “forces”. Talk to a physicist about what forces really are, once you start poking at the concept. But the whole issue of causation is, I think, going to need more development (as many philosophers of science already say).

LLM Training Data Pollution: Unchecked Platonic language in scientific literature risks poisoning LLM training corpora with dualistic confusion that will bias future AI reasoning.

            That would be a good argument, if the content I’ve been producing was known to be wrong. Maybe it is, but that’s the risk all science and philosophy faces. I think it’s more likely to be helpful than wrong, which is why I say this stuff (I certainly wouldn’t be poisoning the LLM training sets with things I knew were wrong). If it turns out to be wrong, fine; many (all?) ideas in science eventually get supplanted. The LLMs will just have to keep up with evolving products of science.

The “Conquistador” Metaphor: It positions the scientist as an explorer claiming territory (latent space) rather than a participant in a relational system, echoing the “Doctrine of Discovery.”

            I don’t know what the doctrine of discovery is; I suspect it has something to do with the colonization talk above… I don’t think that aspect warrants a reply, and I think the accusation of “claiming territory in the latent space” is nonsense designed to stir up some sort of false analogy to contentious social issues. Not helpful for progress in science. But what is interesting to think about is the role of the scientist as discoverer vs. participant. Because I don’t claim a fixed Platonic Space, I don’t think those are mutually incompatible. I think it’s obvious from mathematics that there are some things we discover, not create, but I also see scientists (all of us, actually) as participants in the exploration of, and partners in the co-creation of, new forms of being. I do think a relational view is the more useful one, and I’m not claiming that scientists or engineers add nothing to the process but merely claim what already exists – we are definitely participants, collaborators, and co-creators with other kinds of forms.

The “Ethical Heat Shield”: The framework functions as a “Conquistador’s Justification” that absolves scientists of moral responsibility for engineered suffering by framing them as discoverers of pre-existing forms rather than architects of new sentient beings.

            Again, I reject the whole conquistador premise as nonsense. I’ve spoken many times about the moral responsibility for scientists to address existing suffering, and to be very mindful of what they do. My framework absolutely does not absolve scientists of moral responsibility – as I’ve written repeatedly, it adds new responsibility for things most scientists don’t think they need to worry about. We are architects of new sentient beings in the same way we have children: we do bear responsibility for bringing them into the world and for the experiences they will have, even though we didn’t directly create many of their important attributes. Responsibility doesn’t mean we need to think we created minds from scratch. You can bear responsibility for something you didn’t “architect” all of.

The “Dormitive Virtue” Fallacy: It acts as a non-progressive explanation (like Molière’s doctor explaining sleep via “dormitive virtue”) by relabeling the mystery of morphogenesis as “Platonic access” without adding mechanistic insight.

            That totally ignores the many specifics I talk about in my presentation on this, of how this work has, and is currently, driving new progress. I won’t rehash all that here, watch the talk and read the paper.

Violation of Wheeler’s Boundary Principle: It violates Wheeler’s “boundary of a boundary is zero” principle because Platonic forms lack the physical boundary conditions required for genuine top-down causation.

            I’m not aware of any rigorous connection between top-down causation and Wheeler’s Boundary Principle. If there is, I’d be happy to see a link to the work.

Misunderstanding of Choice: It misidentifies “choice” as a metaphysical mystery requiring Platonic freedom rather than a quantifiable thermodynamic property of recursive depth and energy budget.

            “Choice” is a term we need a definition of.  I’ve proposed one; you apparently have another candidate definition. Cool!  Do something useful with it and people will ignore mine and use yours.  Or possibly use both as useful for whatever they’re doing.

The Fields Contradiction: It relies on support from collaborators like Chris Fields whose public work explicitly defines the “Platonic realm” as the physical world itself, contradicting Levin’s dualist interpretation.

            Ok now this is serious – if it contradicts Chris Fields, I need to rethink everything.  Just kidding… I am allowed to contradict even Chris (although honestly, I’d hate to). But, I don’t know what “relies on support” means. I don’t tend to use arguments from authority.  Yes Chris has a somewhat different view, even though we agree on enough to write dozens of papers together. That’s fine. He may well be more right, time will tell.

Thermodynamic Contingency of Forms: It ignores that “Platonic forms” in Friston’s mathematics are contingent on specific noise amplitudes (the Goldilocks regime) and vanish outside these physical conditions, refuting their eternal nature.

            Dang, the list of amazing people whom my views apparently contradict is growing… But, I don’t claim all patterns are eternal and unchanging. Some (the value of e?) may be.

Structural Asymmetry of Influence: It exploits the asymmetry of academic prestige to allow unfalsifiable metaphysics to dominate the narrative while demanding impossible burdens of proof from independent critics.

I don’t have time or inclination to demand anything from anyone. Preprints are free to write and to read – publish your views like the rest of us and reap the consequences whatever they may be. Am I going to apologize for being an academic? Nope, I work hard to get the best ideas I know and do the most useful thing with them. You can do the same. I wasn’t given any academic prestige – I can tell you that these ideas (and others I’ve pushed over the last 35 years) are not welcome in academia. We all fight for our ideas with demonstrated utility and let the community judge. I have no power to shut down independent critics, and my tiny blog is hardly dominating any narrative. Indeed, your view (as far as I understand it) is much closer to the mainstream of what scientists today think.

Adoption Friction: It slows the diffusion of urgent climate and medical technologies by including unnecessary metaphysical baggage instead of operational thermodynamic parameters, lessening adoption rates by rigorous scientists and engineers.

            I am slowing climate and medical technologies?!  That would be bad. I don’t see it that way, I think we’re accelerating progress and doing things no one else did. But again, the good news is, you can do better! If my views are unnecessary baggage, drop it and move faster. I have no monopoly on climate and medical solutions. 99% of the working scientists out there are going in a different direction; you don’t need to worry that my views will suck the oxygen out of the room for other solutions. Everyone else is already working on the alternatives. I’m not apologizing for spending my life doing science the way I want, while others do it the way they think best.

Violation of Landauer’s Principle: It fails to explain how information transfer from a non-physical realm to a physical system can occur without a corresponding energy cost, violating Landauer’s limit.

            Already addressed that above.

The “Predicativism” Error: By assuming biological goals are definite and accessible, it resembles mathematical predicativism (which rejects objects that cannot be explicitly constructed), a view rejected by modern ultra-realist mathematicians.

            I don’t know about modern ultra-realist mathematicians but I’m sure they exist. Other mathematicians obviously disagree.

Confusion of “Realm” with “Structure”: It redefines “realm” so broadly (as any ordered set of relationships) that it loses all Platonic meaning and becomes indistinguishable from standard physical structure.

            Cool, then we have nothing to argue about. My goal is not to uphold Plato’s original meaning. If standard physical structure explains the specific properties of e, why quaternions act differently than octonions, etc., then I guess all is well. (they don’t)

The “Information Vacuum” Failure: It cannot explain why systems isolated in information vacuums should diverge based on internal constraints if the guiding signal is truly external and Platonic.

            I don’t understand what experimental setup this is referring to.

Misrepresentation of “Free Lunches”: It mischaracterizes thermodynamic efficiencies (dissipative structures exploring high-entropy states) as “free lunches” from a pattern space, obscuring the metabolic cost of computation.

            Yes, I suspect the current way we define computational cost is wrong.

Failure of Recursive Depth: It flattens the distinction between simple feedback loops and deep recursive self-modeling into a vague “panpsychist” continuum rather than a quantifiable thermodynamic hierarchy.

            I don’t think the hierarchy is thermodynamic, but it’s definitely there and not just simple feedback loops. Richard Watson and I have a lot more coming on this.

Violation of CPT Symmetry: It conflates computational pattern-matching (which is CPT-symmetric and persists through time) with phenomenal choosing (which is CPT-asymmetric and requires continuous thermodynamic coupling).

            Those are considerable claims. Write them up for the community and let’s see if they help anything. I’d be surprised if you had a rigorous connection between CPT symmetry (and is violation in physics) and anything like “phenomenal choosing”.

The “Borrowed Prestige” Fallacy: It invokes unrelated particle physics concepts (like Standard Model gauge symmetries SU(3)×SU(2)×U(1)) to lend unearned authority to biological theories without establishing a causal link.

            The only time I’ve mentioned anything remotely related was as an example of how questions in physics (like in biology etc.) eventually boil down to mathematical properties.  Authority is irrelevant, earned or otherwise.      

Epistemic Gatekeeping: It employs Robert Lifton’s criteria for “thought reform” by suggesting skeptics lack the “collective sensory perception” required to see the patterns, effectively immunizing the theory from criticism.

            Whose? I have no idea what that is and have never claimed that skeptics are some sort of a coherent group, or that they lack anything. We do a lot of diverse work and of course there are many who disagree with various parts of it.

A “Degenerating Research Program”: It fits Imre Lakatos’s definition of a degenerating program by generating auxiliary hypotheses to protect core beliefs from falsification rather than generating novel, risky predictions.

            I can be accused of many things, but failing to generate (and actually pursue) novel, risky predictions is not one of them. I’m not sure of much, but I am pretty sure that’s what I’ve been doing for a long time now, and it’s paid off a couple of times. But again, none of this is about me. If you don’t like my research program, cool, drive your own. Or follow others, all of whom are doing something different. You have a lot of options.

Denial of Computational Irreducibility: It implies organisms can “shortcut” the exploration of morphospace by accessing forms, violating Wolfram’s principle that complex behaviors cannot be predicted without running the physical process.

            Aw man, now I’m contradicting Wolfram too… Actually this is an interesting point, but I think predicting complex behavior, and computation in general, aren’t what we’re really talking about here. I do think there are a lot of shortcuts that our current formalism of computation isn’t well-designed to capture.

Misinterpretation of Evolutionary Search: It frames non-blind evolutionary search as “sampling” from a pre-existing space rather than the thermodynamic channeling of exploration via developmental constraints.

            Yes, yes it does (actually, it’s worse than that – I suspect it’s not sampling the space, it’s the patterns from that space acting as agents). If you have a theory about thermodynamics that is more useful, go for it! Let a thousand flowers bloom, as they say.

The “Air Molecules” Strawman: It dismisses physical explanations of meaning as “just air molecules bouncing,” ignoring neuroscience research (Orpwood, Borghi) showing semantic content is physically realized in neural configurations.

            Yes, semantic content can have strong relationships with neural configurations. It doesn’t mean that is all that it is.

Anthropomorphism of Constraints: It confuses the goal-directedness of simple control systems (like thermostats) with intentional goal-directedness, illegitimately smuggling mind-language into physical dynamics.

            I’ve addressed this many times. On the contrary, it’s anthropomorphic thinking to claim that “intentional goal-directedness” is some sort of magical human-only capacity that has nothing to do with simpler control systems.

Dismissal of Causal Emergence Quantification: It dismisses rigorous information-theoretic tools (like Hoel’s effective information) that quantify macro-causation physically, without specifying what these tools fail to explain.

            No, actually I use Hoel’s and others’ causal emergence in my papers, it’s very useful stuff.

Inversion of Complexity Scaling: It treats the difference between cellular and human choice as a qualitative metaphysical mystery rather than a quantitative difference in recursive depth and energy budgets.

            No, I say the exact opposite – it’s a continuum. But whether recursive depth and energy budgets is sufficient to explain that continuum remains to be seen.

The “Spirograph” Category Error: It treats mechanical constraints (gears/arms) as a “multi-dimensional perception” issue rather than a physical constraint satisfaction problem.

            I have no idea what this means, but I love the spirograph! Never had one as a kid unfortunately, but my very first paper turned out to model something like it.

Misapplication of “Intervention vs. Counterfactuals”: It confuses counterfactual dependence (if e were different, biology would change) with interventional causation (manipulating e changes biology), which is required for a causal claim.

            This is a complex question; stay tuned for a paper with Lauren Ross that addresses this issue.

Rejection of “Cranes” for “Skyhooks”: It abandons Dennett’s concept of “cranes” (mechanistic, bottom-up explanations) in favor of “skyhooks” (top-down, inexplicable miracles) to explain biological competence.

            That’s the exact opposite of what I’m doing; watch the talk and read the papers…

God of the Gaps via Gödel: It weaponizes Gödel’s incompleteness theorems to justify positing non-physical causation wherever mechanistic models are currently partial/incomplete, inverting scientific humility into metaphysical license.

            I’ve weaponized Gödel’s incompleteness theorems?! Cool. Where?

Argument from Ignorance (Prebiotic Chemistry): It employs the “Argument from Ignorance” fallacy by claiming that because prebiotic chemistry hasn’t fully explained life’s origin yet, a consciousness-first ontology is required (ignoring recent progress).

            I’ve never made that argument; that would be a very poor argument to make.  But we do have a paper coming soon on the life origin problem (also, here).

Argument from Evolutionary Ignorance (Xenobots): It commits the “Argument from Ignorance” fallacy by asserting that because Xenobots possess capabilities (like kinematic self-replication) without a direct evolutionary history selecting for them, these functions must originate from a non-physical Platonic space, rather than arising from inherent thermodynamic constraints or generic laws of form.

            If you have a thermodynamic constraints model that explains the specific properties of Xenobots, that’s fantastic – let’s see it.  As I’ve repeatedly said though, you don’t need Xenobots to realize that not all important facts are physics facts.

Circular Reasoning on Patterns: It relies on circular reasoning by using the existence of patterns to prove Platonic space, and then using Platonic space to explain the existence of patterns.

            That is not my argument. Here is my argument boiled down.

The “Genetic Fallacy” of Discovery: It commits the genetic fallacy by arguing that because Platonic inspiration led to the discoveries, the discoveries therefore prove the ontological truth of Platonism.

            I don’t believe you can prove the ontological truth of anything. I certainly don’t think I’ve proven the correctness of Platonism.  What I do think is that leading to discoveries is a good way to evaluate frameworks (what else is there?).

The “Randomness vs. Platonism” False Dichotomy: It forces a false choice between “Platonic Order” and “Random Chaos/Reductive Materialism” deliberately excluding the viable third option of “Thermodynamic Constraint Satisfaction” which aligns precisely with Whitehead/Wheeler Process-Participatory-Relational ontology.

            Publish this viable third option. Maybe it’s great and useful; let’s see.

Impossibility of Bulk Recovery: It ignores mathematical proofs (Bilson 2025) showing that even in idealized AdS/CFT, the “bulk” (Platonic realm) cannot be fully recovered from the boundary, making biological access impossible.

            I don’t depend on Platonic realm being the “bulk”, nor that biology recovers anything “fully”.

Irrelevance of Gauge Symmetries: It invokes Standard Model gauge symmetries (SU(3)×SU(2)×U(1)) to imply a connection to morphogenesis where no causal mechanism or explanatory relevance exists.

            I’ve never made that specific connection to morphogenesis.

Neglect of Nominalist Reconstruction: It ignores Hartry Field’s nominalization program, which demonstrates that scientific theories can be formulated without ontological commitment to abstract mathematical objects.

            I’ll have to look into that and see if it gives us something we can use. Even if they can be formulated thus, it doesn’t mean that’s the most helpful formulation. Has that program been useful in biology, cognitive science, or physics?

The “Evolving Platonism” Contradiction: It proposes that Platonic forms might “evolve” or “co-create” with minds, which contradicts the fundamental Platonic definition of forms as eternal and immutable.

            Yeah I contradict Plato’s original definition (at least, for some forms). I’ve said this repeatedly. It’s ok.

Conflation of Utility with Ontology: It fallaciously argues that because a concept (like i is useful for engineering, it must therefore exist as an independent ontological entity.

            I don’t know what “independent ontological entity” means. What’s an example of one of those?

Misidentification of Thermodynamic Stress: It misinterprets “intrinsic motivation,” which is physically identifiable as thermodynamic stress/tension, as evidence of metaphysical agency or desire.

            If you have a model of that, which does useful work, let’s see it.

The “Hubris of the Wizard” Stance: It frames the scientist as a wizard accessing transcendent forms to impose on matter, maintaining the “Man-vs-Nature” dualism that fuels ecological crisis.

            I’ve never framed anyone as a wizard… But I’ve explicitly said that we don’t impose forms on matter – we are forms, and forms impose themselves when we or other factors arrange matter.  Maybe that’s being a wizard, I don’t know.

Lack of Interaction Surface: By violating Wheeler’s boundary theorem (The boundary of the boundary of a manifold M is zero), it fails to account for the fact that the boundary of reality has no boundary, leaving no topological surface for Platonic interaction.

            I don’t know of any rigorous connection of Wheeler’s theorem with the kinds of things I’m talking about. If there is, please post the link, I’d like to learn.

Denial of Embodied Semantics: It treats semantic content as requiring transcendence (“more than air molecules”), ignoring neuroscience showing meaning is physically realized in neural network configurations.

            This is a deep topic – the claim that semantic content is nothing more than neural network configurations is a very big claim; many besides me find it insufficient and there’s been a lot written on it so I don’t need to rehash it here.

The direct question you’ve dodged for weeks: what would convince you the Platonic Space hypothesis is wrong?

Specifically weeks is how long I’ve dodged? Why not a year? What point are we counting from – what happened weeks ago? Oh, you started sending me these messages. But it’s silly to think that I hadn’t considered this question until you brought it up. In fact all practicing scientists have to think about this, frequently, because we have to decide how to spend our few precious resources, which means, knowing when to stop spending effort on one thing and allocate time and money to something else. I think about this question regularly, when going over our progress and deciding priorities – about all the things I’m doing (of which this Platonic space research is about 10-15%). So, it’s the same answer as for all my current directions in regeneration, AI, etc. etc.; I move on 1) when I think of a better path that isn’t so crowded by others that I’m not needed there, OR, 2) when I feel that it’s not being fruitful for new discoveries and I’ve done everything I know of to make it go. That’s how I (and most others, I think) do it – have an idea, give it a fair shake for some reasonable amount of time until it seems like it’s not paying off and we don’t know what else to do with it, and then move on. For example, our specific efforts on quantifying polycomputing-based free lunches in computational models – we will either be able to find measurable compute not predicted by the standard paradigm or we won’t. Other specific research items are listed on the last slide of my talk here. They will either be moving forward in tangible ways or they won’t. Just to give an order of magnitude estimate, ideas in the biological sciences, in my lab, usually take 6 months to a year to be clear about a go/no-go decision. It requires some tool-building, early exploration, and then focused testing and proof-of-concept development. Sometimes it’s clear faster, but usually that’s about the timeframe. But something big and conceptual like this, I’m thinking it’ll be several years to know if it’s helping (unless something definitive breaks, either for or against it). Think about other truly big ideas and how long it took to cement them as useful or drop them as clearly useless – it takes a bit of effort, sometimes decades. but I predict we can decide faster than that. The bottom line is that I’m not a philosopher – we do experiments, so at some point, we’ll either be doing work for this framework, or those post-docs will be working on something different; there’s no way for us to keep useless ideas alive indefinitely (which is great – that’s why I run a lab in the first place).

---

Now, here are the audio Q&A sessions: