Donald Hoffman Meets Stephen Wolfram For the First Time on TOE

ELI5/TLDR

Two of the more unusual thinkers in science — a cognitive scientist who argues spacetime is basically a video-game interface our minds made up, and a computer scientist who thinks the universe is a giant rulebook being run out from every possible starting point — sit down for three hours and try to figure out whether they’re saying the same thing in different languages. Donald Hoffman wants consciousness to be the bedrock of reality; Stephen Wolfram wants computation to be the bedrock. They spend the conversation circling a surprising discovery: their two bedrocks might actually be the same object viewed from opposite ends. An earthquake interrupts them near the end, which they handle with the calm of two people who have already decided reality probably isn’t what it looks like.

The Full Story

The setup is almost a joke. Curt Jaimungal, who runs a podcast called Theories of Everything, has gotten Donald Hoffman and Stephen Wolfram on a video call for the first time ever. Hoffman is the Kant-flavored cognitive scientist who argues that the chair you’re sitting on doesn’t exist when you’re not looking at it — that space and time are a kind of user interface our consciousness built, not the real thing underneath. Wolfram is the guy who built Mathematica, left physics for decades, then came back and announced he’d found the actual low-level machine code of the universe. Put them in a room and either one of two things happens: they talk past each other, or they discover they’ve been climbing the same mountain from opposite sides.

Mostly the second thing.

Hoffman’s opening gambit: show me the taste of mint

Hoffman starts with a trap he’s sprung on a lot of neuroscientists. Current theories of consciousness — the prestigious ones, with grant money behind them — claim consciousness arises from the right kind of physical activity in the brain. Integrated Information Theory says you need a certain “causal architecture.” Global Workspace Theory says you need the right kind of broadcasting. Roger Penrose thinks quantum vibrations in little brain structures called microtubules are doing it.

Hoffman’s challenge to all of them is the same. Humans can have something like a trillion distinct conscious experiences — the taste of chocolate, the smell of garlic, middle C on a piano. If your theory is correct, just pick one. Any one. Show me the specific physical configuration that produces the taste of mint. Not a causal architecture in the abstract — the actual matrix, the actual numbers. How big is it? What’s in it?

“We’re interested in scientific theories that explain specific conscious experiences. What experience can you give me? Humans can experience around a trillion different experiences, so it should be like shooting fish in a barrel. Which ones have you done? The answer is zero.”

This is the gap Hoffman thinks nobody has closed. And he doesn’t think anyone can, because he thinks the whole project is backwards.

The headset metaphor

Here is Hoffman’s actual view, which sounds strange and then becomes coherent if you sit with it. When you look at another person, you’re not seeing them. You’re seeing pixels on a screen. The pixels are your mind’s interface to something else — some pattern of conscious experience that lives beyond the interface. Your body is a pixel. A frog is a pixel. The atoms and neurons physicists study are pixels. Space and time themselves are pixels. The real stuff is consciousness, full stop, and four-dimensional spacetime is just the particular headset our kind of mind uses to interact with it.

The best analogy here: think of a VR headset. Inside, you see a world with physics and objects. None of it is actually in the headset. The headset is a window onto computation happening elsewhere. Hoffman is saying: our experience of physical reality is exactly like that, and our mistake is assuming the stuff inside the headset is the real thing. Other minds, he suspects, use very different headsets — some with a billion dimensions instead of four. Our three-plus-one-dimensional universe is one of an infinite number of possible interfaces, and probably not an especially sophisticated one.

The reason he holds this view, scientifically rather than mystically, is that he thinks the alternative is logically impossible. You cannot, he argues, get mind from non-mind. You cannot take unconscious gears and have consciousness emerge from their grinding. This is the old Leibniz argument — Leibniz had a thought experiment where you imagine walking inside a giant mill that is somehow producing thoughts, and you look at all the gears, and no matter how closely you examine them, nothing about the gears will ever explain why there is an experience of anything. Leibniz thought this argument was so obviously right he only gave it one paragraph.

Hoffman thinks Leibniz was right. The way out, for him, is to put consciousness in the starting ingredients rather than trying to cook it up from non-conscious stuff.

Wolfram’s counter: maybe you’re underestimating what simple rules can do

Wolfram is sympathetic but not buying the “logically impossible” part. And his objection comes from forty years of actually running computer experiments on simple systems.

Here is the thing Wolfram has learned that most people haven’t. You can write down a set of rules so simple they look boring — like, “at each step, look at a cell and its two neighbors and apply this three-bit lookup table” — and when you run them, extraordinarily complex, lifelike behavior emerges. Wolfram’s famous example is Rule 30, a cellular automaton with basically nothing going on at the level of the rule, which produces patterns so complicated they are indistinguishable from randomness and can even be used as a random number generator. He spent three years just refusing to believe what his own computer was showing him.

The point: Wolfram’s prior, built from decades of experiments, is that richness pops out of sparseness constantly, and that our intuition for what “too simple to produce anything interesting” looks like is reliably wrong. So when Hoffman says you can’t possibly get consciousness from non-consciousness, Wolfram’s response is basically: I’ve heard that tune before, and it usually turns out the singer just hadn’t run enough steps of the simulation.

Wolfram’s own picture of reality is something called the ruliad. Imagine you took every possible rule — every way a universe could conceivably evolve — and you applied all of them, to everything, forever. The ruliad is that entire tangle. It is assumption-free in the sense that it contains literally every possibility. Physics, in Wolfram’s view, is what we see when we — specific kinds of observers with specific limitations — take a thin slice of the ruliad and try to describe it. The reason we see general relativity and quantum mechanics isn’t because those are fundamentally true. It’s because those are the laws that any observer sufficiently like us must see.

And the two critical features of observers like us that generate our physics: we are computationally limited (we don’t get to track every detail, only aggregated features), and we believe we are persistent over time (we experience a single thread of “me-ness” rather than a branching cloud of parallel selves).

The accidental convergence

Here is where the conversation gets interesting. Hoffman, starting from consciousness, has built his own mathematical structure. He calls it conscious agents — basically a giant network of minds linked by Markov chains, which are just mathematical objects that describe probabilities of moving from one state to the next. (“If I’m seeing red now, what’s the probability the next thing I see is green?”) He and his team have worked out a partial ordering on these things — a way of saying one conscious agent “contains” another by being a richer version that can be collapsed down to the smaller one by averaging over details.

Wolfram listens to this and gradually starts recognizing the shape. He asks a series of sharpening questions. Are the matrices finite? Does the structure have a top? (No — Hoffman says you can go infinitely far in infinitely many directions.) Do you have products of these kernels, or just tracing? (Products too, eventually.) And then Wolfram delivers the punchline he seems genuinely surprised by:

“I think what you’re constructing — if you replace the Markov chain thing with arbitrary computation, you basically have the ruliad. You have the same object.”

In other words: Hoffman is building from “consciousness” upward, Wolfram is building from “computation” downward, and they may be converging on the same mathematical object from opposite directions. Hoffman’s partial order of consciousnesses and Wolfram’s ruliad might be — give or take some technical cleanup — the same thing.

Hoffman’s response is essentially “delighted, if true.” And then he pushes back on the idea that they’re equivalent by noting that Markov kernels are computationally universal — meaning, in principle, you can build any computation out of them — so if there’s a relationship, he thinks it should be full equivalence, not a subset. Wolfram pushes back on that pushback (probabilistic things and purely computational things aren’t quite the same flavor, he says). But the family resemblance is striking enough that both of them keep circling back to it.

The God question, politely

Near the end, Curt tries to summarize. He points out that Hoffman, pushed on how one can have multiple consciousnesses when all you really know firsthand is your own single experience, had earlier said something like: maybe there’s just one infinite consciousness, and “you” and “I” are different views it takes of itself. Different straws through which the same whole looks at itself.

Wolfram notes, drily, that this infinite consciousness thing sounds a lot like the ruliad. Curt notes, also drily, that if you take Hoffman’s “all I intimately know is my own consciousness” seriously and run it to its limit, you end up claiming you are God.

Hoffman doesn’t flinch:

“I’m willing to go there. But I’m taking you with me. I’m saying that you and I are both God looking at self, talking through two different avatars.”

At which point an actual earthquake hits Hoffman’s house in Southern California. Wolfram, without changing tone, observes that they must be getting closer to the truth. Hoffman confirms the cats are fine.

What they actually agree on

Strip away the vocabulary and there is a surprising amount of common ground:

• Spacetime as we know it is probably not fundamental. Both men think it emerges from something deeper.
• The reason we experience physics the way we do is tied to what kind of observer we are. A different observer would see different laws.
• The universe has way more going on than we can perceive. We are sampling a very thin slice of something much larger.
• Probability, when it shows up as a foundational ingredient, is usually a sign that your explanation has stopped and something deeper is needed.
• The experience of consciousness is hard to communicate because we haven’t figured out how to package it in a form that survives the trip from one brain to another. Language is our best trick for this and it’s still pretty bad.

And the disagreements are mostly about ordering. Hoffman insists consciousness has to be in the ingredient list from step one. Wolfram insists that’s a move you make because you’ve given up on cooking it from cheaper ingredients, and he has not given up.

Claude’s Take

This is a genuinely high-quality conversation, by which I mean: both participants are actually listening, actually updating, and actually disagreeing about things that matter rather than performing disagreement for the camera. That is rarer than it should be.

Hoffman’s “show me the taste of mint” challenge to physicalist theories of consciousness is legitimately sharp, and the physicalist camp really does struggle with it. There is no current theory that can point to a specific physical configuration and say “that is the taste of mint.” That’s a real gap. Hoffman is not making this up.

However. The move Hoffman makes from “no one has done it” to “it is logically impossible” is much weaker than he lets on. Wolfram catches this and pushes on it, and Wolfram is right to push. “Forty years of experimenting with simple systems taught me my intuitions about emergence are systematically wrong” is exactly the kind of empirical calibration you want from someone weighing in on what can and can’t arise from what. Hoffman’s certainty here is essentially Leibniz-flavored intuition, dressed up in more recent math. Intuition is a perfectly reasonable starting point for a research program — it is not an argument that closes the case.

The interface theory itself (“spacetime is a headset”) is philosophically defensible and taken seriously in some corners of cognitive science, but the leap from “our perception of reality is not the same as reality” — which is obviously true, your phone is not literally a glowing rectangle of pixels to the universe — to “therefore consciousness is the fundamental stuff” is a very big leap that does a lot of unearned work. You can believe the first without being forced into the second. Hoffman sometimes talks as though the second follows from the first. It doesn’t.

Wolfram is on stronger ground when he stays in his lane. The ruliad is a weird, fertile idea, and the claim that you can derive general relativity and quantum mechanics from very minimal assumptions about observers is genuinely novel work that physicists like Nima Arkani-Hamed have taken seriously. The claim that spacetime is emergent from something more fundamental is currently one of the more exciting research directions in theoretical physics, and Wolfram’s framing is one of several plausible versions of it. Not proven, not mainstream, but serious.

Where Wolfram gets softer is when he claims “I think we got it, I think we know how it works.” That’s a very Wolfram move — historically his calibration on “I’ve solved the big problem” has been poor. A Wolfram model that successfully derives the specific structure of the Standard Model of particle physics from his hypergraph rewriting rules does not yet exist. He keeps saying it’s around the corner. It has been around the corner for a while.

The part where they both start recognizing that Hoffman’s network of conscious agents and Wolfram’s ruliad might be the same mathematical object is the most interesting moment of the conversation, and also the part most likely to evaporate under closer scrutiny. Both structures are doing similar jobs — serving as a kind of universal substrate that you project down from to get a particular observer’s perspective on reality — and there are legitimate mathematical reasons why, if you generalize far enough, those kinds of constructions tend to converge on each other. But “family resemblance at the limit of generalization” is not the same as “equivalent structures.” Hoffman’s Markov kernels, probabilistic by construction, and Wolfram’s rule-based computational universe are doing different things in their respective bones, and some of those differences really matter (this is the whole point of Wolfram’s insistence that probability is not the same as non-determinism — a technical point he is correct about). The equivalence, if there is one, would need to be proved, not gestured at during a podcast.

Hoffman’s specific mathematical machinery — using decorated permutations and positive geometries to project consciousness dynamics down into spacetime — is at the edge of contemporary theoretical physics. That edge is real. The amplituhedron is a real thing, the European research push into “positive geometries beyond spacetime” is a real thing, and the combinatorial classifiers (decorated permutations) are real. Whether Hoffman’s bridge from his conscious-agent networks to those structures actually works is an open question and one he himself admits has not been demonstrated. The plan is credible; the execution remains unfinished.

Finally: the move at the end, where Hoffman is willing to say “you and I are both God looking at self through two avatars” — this is coherent with his framework. It’s also the place where the framework stops being science in any operational sense and becomes metaphysics of a kind that is unfalsifiable by construction. There’s nothing wrong with that, but it should be labeled accurately. Hoffman mostly does label it accurately; he is honest that what he’s offering is an ontology and that its scientific payoff depends on whether the math can eventually produce testable predictions about particle physics. That is an entirely reasonable bet, and the bet has not yet paid off, and he knows it.

The conversation is worth watching if you find it interesting to see two people whose priors are almost diametrically opposed discover, in real time, that they might be describing the same elephant from opposite ends. The earthquake is just a bonus.