Is the Brain an Interface?

Michael Levin said something on Lex Fridman's podcast that made me stop the video and sit with it for a while. What if a brain is an interface to mind, rather than the place where mind begins?

The thin-client analogy came easily to me. A terminal can display something whose source is elsewhere. But an analogy that comes easily deserves some suspicion, especially when it happens to fit a philosophy I already find compelling.

Levin's proposal about a “Platonic space” of patterns is a speculative extension of his research. The biological experiments and the metaphysical interpretation carry different kinds of evidence. I want to keep that distinction visible while following the idea. The conversation is the starting point for this essay.

The step that interests me

Levin begins with mathematical patterns. Their properties constrain what physical systems can do, even though changing a physical object does not rewrite arithmetic. He then asks whether the space of possible patterns could contain richer, more agent-like structures, with bodies providing ways for those structures to appear.

That last step is substantial. The existence of mathematical regularities does not, by itself, establish a reservoir of minds. Calling a brain an interface gives us a picture to think with. It does not yet tell us what the brain interfaces with, how the connection works, or how we would distinguish it from a brain producing the same behaviour through physical processes.

I find the picture interesting because it changes the question I am inclined to ask. Instead of assuming that the familiar human arrangement is the only useful starting point, it invites a wider search for organised behaviour. The risk is that the picture becomes so accommodating that every possible result seems to confirm it.

The resemblance to Spanda

Kashmir Shaivism offers a different starting point. In the Spanda tradition, consciousness has a dynamic character. “Vibration” here is not a claim about an electrical frequency in tissue. It is a way of describing the activity of awareness and manifestation. The Lakshmanjoo Academy's account of the schools explicitly distinguishes this movement from merely physical motion.

That is the association I heard in Levin's account: a particular embodiment need not be the ultimate source of everything expressed through it. I can recognise that resemblance without treating the two accounts as interchangeable.

The traditions are doing different work. Kashmir Shaivism concerns recognition, experience and liberation. Levin is proposing a research direction in which patterns and the bodies that realise them might be mapped. A useful experiment should tell us something that a competing account would not have led us to expect. A contemplative insight is assessed differently. The same word, consciousness, does not make their methods equivalent.

I am drawn to the resemblance, and I want the freedom to ask a strange question without knowing in advance where it will lead. The resemblance is a reason to read more carefully, not a result we can announce.

A similar name, another claim

There is also a machine-learning paper called the Platonic Representation Hypothesis. Huh and colleagues investigate alignment between the representations learned by different models. Their proposal concerns shared statistical structure in data and learned representations.

It is easy to put that paper beside Levin and feel three different conversations clicking together. But similarity between model representations does not demonstrate disembodied minds, and it does not settle whether any model has subjective experience. The shared adjective is doing more work than the evidence if we use it that way.

What would make me take the interface seriously?

I would want the proposal to constrain our expectations. Which capacities should appear in which physical arrangements? Which should be absent? Can it predict an outcome before we build the system, rather than give that outcome a new description afterwards?

There is room for philosophical speculation before those answers exist. It can direct attention toward experiments nobody thought to try. I would just like to know when I have crossed from the experiment into the interpretation.

The conversation left me with a better question than “did an ancient tradition already know this?” If the brain is an interface, what would we expect to observe that we would otherwise miss?