I have been thinking about the insufficiency of physicalism and its scientific and personal implications for somewhat over 40 years. But I’ve only started talking about them publicly in 2025, because I saw no point in adding to the Philosophy and New Age bookshelves – relevant ideas are ancient, and plenty of much smarter people have opined in this direction without moving the needle much on the mainstream paradigm. What’s different now however is that these ideas have become more strongly actionable. Not merely “testable”, but able to drive new research programs. Thus, it becomes important to lay out a conceptual foundation that is informed by the latest discoveries, and sketch a research program that this framework enables. It’s time, and I’ve now mentioned this in a number of talks.
There is currently a convergence of work in this direction from others as well. The machine learning (computer science) community is approaching it from one side. The cognitive science field has its own “anomalous” data on the linkage between mind and brain which suggest there may be foundational gaps in our understanding (see one review here). And of course, non-physicalist theories in philosophy abound. What I’m doing is taking an approach that I think is new in 3 ways: I’ve tried to
(1) anchor it in experiments in developmental/synthetic biology, and interface tightly with the diverse intelligence field,
(2) make a specific proposal for understanding causation and interactionism (the claim that the mind:brain relationship is the same as the math:physics relationship) that connects well to Platonist ideas in mathematics but extends them far outside of math, and
(3) broaden the usual dualistic framework, in ways that neither materialists nor organicists tend to like by arguing that the non-physical patterns that drive advanced cognition are drawn from the same space as those that drive minimal physical and computational systems. In other words, that there are no truly dumb machines and dead matter, nor any magical divide that makes us unique in being more than our formal models of algorithms and physics.
The long-form version of my current views can be see in this blog post and videos (this, and this briefer one). Also the preprint (although it’s now quite dated and is being revised and improved to a new version); much more is coming on this in the form of papers. Of course, there’s also our symposium which has many others’ excellent, related talks.
Here, I wanted to present the bones of the ideas in a tight way that attempts to make the logic clear and simple. There’s a lot of misunderstanding of the claims (for example, some people think that I’m going down this road only because of Xenobots etc.; that is not the case). It’s helpful to lay the sequential argument out step by step so everyone can see what it leans on, and perhaps decide which step (or which inference) they disagree with. The left column of the table is close to a philosophy/logic format of argument. The right column is some commentary on each one. Here goes:
| Argument | Commentary |
| There are facts which are not physical facts. | These include truths of topology, the shape of specific fractal structures, the distribution of primes, the contents of your discrete math and computer science textbooks, the fact that complex numbers, octonions, quaternions all keep different promises, etc. They cannot be found using the tools of physics, and knowing physics doesn’t help. Can you disband the math department and hope the physicists find them? No. Can they be changed by tweaking the universal constants at the time of the big bang? No. If you agree that you can’t simply dissolve the math department, then that’s it – we have 2 distinct (but interacting) realms. Now we can do some science and understand how those relate, and how other disciplines come into it. |
| Let’s call these facts (for maximum generality) “patterns”: These patterns are discovered, not created by us. Therefore we can start thinking about the latent space from which specific ones are drawn. | I realize this is a huge debate in the philosophy of mathematics. What I find convincing is that you can start with something like set theory (define set membership and successors) and eventually end up with a very specific value of e, the natural logarithm. You don’t get to choose it, it’s forced on you by very minimal assumptions. The fact that aliens that started with different assumptions would see a different subset of these facts is not problematic – they would simply have access to a different set of discovered patterns. Others argue that these patterns don’t “pre-exist” in some sense but are generated by us. I’m not sure how much of the below would be affected by that; we’d still have to study the space of them, and answer the questions at the end. |
| These patterns are important for physics, biology, and computer science. | Evolution exploits them as affordances it doesn’t need to pay for. For example, once you find a voltage-gated ion channel, you have a transistor which can make logic gates, and truth tables (and facts like the special nature of NAND) are yours for free, you don’t need to evolve them specifically. There are tons of such examples. Also, we know when we paid the computational cost of evolving natural life forms – in the eons of selection against environments. But when was the computational cost of evolving Xenobots and Anthrobots (and all their cool new capabilities) paid? They’ve never existed, so there was never selection for these features. Where do the specific capabilities and properties of novel beings come from? What determines them? Saying they “emerged” at the time you were evolving frogs and humans just basically rips up the whole point of evolutionary theory (a tight predictive relationship between an organism’s features and a past history of selection in specific environments, which explains those features). No one predicted their new transcriptomes, the sound-response ability of Xenobots, the neural healing ability of Anthrobots, from the selection forces and frog/human environment. Waiting until they show up and then claiming emergence is sterile; we need to be able to predict them, which means, mapping the space. |
| These patterns are causal in physics, biology, and computer science because, a) if they were different, facts in the physical world would be different, and b) they serve as best explanations for why the physical world is the way it is. | If you’re worried about interactionism (functionally causal physical:non-physical interactions), too bad – that ship had already sailed in the time of Pythagoras, who knew that non-physical truths control our world. a) they are causal in Judea Pearl’s sense of counterfactual; for example, if the prime number distribution was otherwise, then the cicadas would come out at years other than 13 and 17. b) I think the best sense of causation, e.g., “A caused B”, is if A is the best (most enabling, insightful) explanation for why B, rather than C, happens. In this sense too, non-physical patterns are causal. Whatever your question – in cognitive science, biology, physics – just keep asking “but why?” and you will always end up in the math department. Why do the fermions do this or that? Because of a symmetry of some mathematical object. Why do the cicadas come out at certain years? Because of the distribution of primes. And so on. |
| Therefore, physicalism is insufficient in an important way and we need to investigate this latent space of patterns. | The above points imply that we have to take these non-physical patterns seriously in our explanations and in our engineering (in a sense, all engineering is partly reverse-engineering, as you’re trying to exploit the autonomous capabilities of your chosen substrate). They have consequences for what we investigate and build, and how. |
| Moving science forward requires us to not assume this space of patterns is random but rather is ordered, structured, and amenable to exploration. | This is a metaphysical stance, and I can’t prove it, but I find the alternative terribly pessimistic. I choose positivity. Same with “emergence” – it’s just not useful enough. First, it just means you got surprised (things that seem emergent to one observer, are obvious consequences to a smarter one). Second, it’s just too defeatist to say “there’s no understandable latent space from which surprising properties are drawn; these ‘regularities’ we observe in the world are just random and that’s that.” Third, we still have to say why one pattern “emerged” rather than another, and to whatever extent there’s an explanation, it’s not emergent anymore but has an upstream structure we can understand. If so, there’s our Platonic Space of patterns. |
| Therefore, we can talk about a Platonic or Latent space of non-physical truths. | If you’re worried about adding new “realms”, sorry, I think there’s no way around it and we have to go where our findings take us – no one promised that monism would be guaranteed or useful at an early stage (which is where we are). The good news is that it’s not a commitment to mysterianism – mathematicians have been doing fine this way and what I’m proposing is a research program, not a retreat into an unknowable space. Note also that if you say “the patterns really are just part of physics” then you’re forcing monism by cheating (re-definition) – in no practical way are the discoveries of mathematicians and physics overlapping in methodology. But if you want to make the patterns part of physics by re-defining physics to mean “Everything of importance”, fine. Then I would say the term “physics” is mostly useless, and we really should just talk about its two major branches, which are quite distinct (you get to name 2 sub-realms, instead of realms) and we still need to try to understand how they relate and nothing below changes. |
| We cannot assume that these patterns matter only for math and are irrelevant beyond that. | There is no good reason to limit Platonic hypotheses to mathematics, although that’s where it has mostly stopped in modern science. I’m dropping this axiom, that most everyone likes to take on. Fewer assumptions is better. Let’s examine the consequences of a broader view. |
| The field of diverse intelligence is helping to understand a continuum of intelligence agency all the way from extremely minimal systems through humans and beyond. | See this paper for a deep dive into the spectrum. The key is a unification of systems with respect to their goal-seeking competencies and creative plasticity, regardless of their composition or origin story. The actual level of any system must be determined empirically (can’t be guessed philosophically). In other words, the field of diverse intelligence gives us the tools to think about a continuum (a space) of very different kinds of behavioral competencies, from extremely minimal to very sophisticated, and many that are very alien in the kinds of spaces they inhabit. This is an important conceptual skill to have. |
| I make the hypothesis that patterns will be relevant also to other disciplines well beyond mathematics. | My specific framework proposes a huge space, whose denizens have been classified by us as belonging to different disciplines. From this view, math is simply the behavioral science of a type of pattern amenable to formal models, developmental biology is the behavioral science of cellular collectives navigating anatomical morphospace, etc. Interestingly, many professionals have no issue with Dualism in mathematics, and no one is worried about an immaterial “algorithm” literally controlling the electron flows in a computer; people are only allergic to these kinds of models when we claim the patterns are relevant to biology or cognitive science, and might themselves have higher-order agency. |
| Conclusion: we can think of physical bodies (embryos, cyborgs, computers, etc.) as interfaces for a non-physical space of patterns that ingress through these interfaces to reveal to us specific behaviors. These patterns may be of highly variable agency. We should think of the mind-brain relationship as analogous to the math-physics relationship. | Some of the patterns in that space are low-agency, like some mathematical facts that just sit there (e.g., value of e); some are slightly better, being simple dynamical systems (e.g., Liar paradox oscillators). Others have simple learning capacity as we showed for systems of coupled ODEs. Others are so dynamic and capable that behavioral scientists would recognize their competencies as behavioral propensities; i.e., they are kinds of minds. |
| Therefore, theories about the physical interfaces – like, the thermodynamic cost of computations, evolutionary theory, neural substrate for mind, etc. are importantly incomplete. | Like with all thin-clients, studying the front end interface will be only part of the story, if the important information is on the back end of the server. I suspect almost everything we know is now up for revision, because we’ve been focusing on only one end of the story – the physical end. |
| Therefore, we don’t have to assume these patterns only constrain – they may offer enablements, which should be studied (quantified, predicted, benefitted from). | This is possible because the “no free lunch” commitments are derived from the laws of the physical world. The Platonic Space seems to offer useful patterns for which the physical processes of learning, evolution, etc. do not need to pay (or, pay some, but receive much more than the effort they put in). I think focusing on constraints is very limiting for discovery. I suspect that “physics” is what we call the study of systems that are constrained by these patterns, but “biology” and “cognitive science” (which I think is wider than biology) are what we call the study of systems that are potentiated and enabled by them (i.e., that exploit the patterns, and are indeed themselves patterns manifesting through embodiments). |
| Now we can ask specific questions: | |
| A) Are the patterns unchanging and eternal? | I don’t commit to that. There may be some, like e, but I think most of the interesting ones (like us – we are patterns on this view) are modified by their projections into the physical world (it’s a 2-way interface). The naming is questionable; I chose Platonic Space because the mathematicians would immediately know what I mean and I could extend off of that, but I make no attempts to stick close to Plato’s views (to whatever extent we know). |
| B) Can patterns interact laterally within the space? | We don’t know yet, but possibly. A mathematician thinking about abstract objects may be an instance of one (highly intelligent human) pattern resonating with a lower-agency feature of the space (the mathematical object). Maybe there are other instances; maybe all memories are not stored in the living tissue (nevermind brain) and are partly a reconstruction from prompt patterns in the space. |
| C) Could it have been otherwise? | I have no claims about how the Platonic space came to be, but we are getting data that may suggest it is somewhat fine-tuned to favor intelligence and agency (like the physical universe is seen to be). Stay tuned for that. |
| D) Are the contents of the space sparse or dense? Finite or infinite (and what grade of infinity?). | We don’t know. I have no strong claims on this. |
| E) Isn’t it a problem to think the patterns somehow pre-exist their physical embodiments? | I don’t know how to handle time in this kind of situation yet (interaction between physical events and non-physical patterns), so I won’t claim patterns pre-existed the physical world – it may not be a well-posed question in this situation. But, we do have to say why specific forms are the way they are (e.g., why it’s the 4-color theorem, not the 9-color theorem), which means that the patterns that emerge are from a specific biased distribution (that collapses to one outcome?) and that distribution has to “pre-exist” in some sense (causal, explanatory). |
| F) So what does the Platonic space actually offer – what degree of free lunches? How do we exploit them, in a measurably consequential form? | This is a major part of the research program. We know the space offers static forms (value of e etc.) but perhaps also dynamic behaviors (policies), and possibly actual compute. We’re testing that now, in a range of simple computational systems (which make things easier to quantify) and biological ones (which offer more impressive and sophisticated examples of free lunches). |
| G) How best to study this space and its influence on the physical world? | We need to build interfaces and fearlessly (and creatively) study what unexpected patterns ingress through them that are not well-explained by their history of selection, engineering, or learning from experience. Biology (e.g., synthetic morphology) offers the most sophisticated patterns, but it’s very hard to prove anything in biology, so we’re also making minimal computational models where we can more easily quantify the effort put in and the outcome we observed. |
| H) Any hypotheses about who or what prepared the space, what the largest pattern in the space might be, etc.? Surely this supports ancient belief systems that would be comfortable with non-physical patterns? | Not from me, no. I know of no way to derive any normative claims from what we know about this field so far, or of any relevant data that favor a specific system over others. I do not encourage that kind of speculation from anything I’ve said. I do think this work (not just my lab’s) will drive major changes in areas of science, and also enlarge the scope of scientific inquiry itself. I do not support using that to shoehorn in various agendas beyond the (broad) scientific search for insight where-ever it leads. |
| I) Shouldn’t you abandon this line of inquiry because it might give cover to those who interpret it in favor of specific belief systems? | It’s not our job to make sure that specific systems remain coverless, and I’m not going to avoid talking about what I really think is going on, out of fear that others may draw inappropriate conclusions, or just conclusions that we may not like. I may not be happy about it when that happens, but it’s a risk no matter what we say or do in science and philosophy and has always been true. The solution is not to bend from what we discover, but to better communicate to the wider public, as clearly as possible, what we think our observations do and do not say. Certainly scientific discoveries in physics, neuroscience, and evolutionary theory have been inappropriately interpreted in many ways (giving rise to the meaning crisis, etc.). I will address those issues elsewhere, eventually; I think this work is not unconnected to that, but it’s way too early to say anything concrete about it yet. |
The specific research program suggested by these ideas is in progress in our lab and others (some of it is listed here but developing quickly so that’s going to be out of date continuously). This is all changing rapidly with new findings, so I may soon need to revise some or all of the above. That’s part of the fun of science. For now, the above summarizes my thinking in this area.
Featured image by Midjourney.
Leave a Reply to James of Seattle Cancel reply