I have been asking AI systems hard questions about intelligence for a while now — partly because the Theory of Embedded Intelligence demands it, and partly because the answers tell me as much about the AI as they do about the question. In May of 2026 I sat down with Google’s Gemini and asked it about Karl Pribram. What came back was a tour through some of the most provocative ideas in twentieth-century neuroscience, with a payoff that lands squarely inside TEI: the brain is not where you think it is, and intelligence is not what you think it is.
What follows is an edited account of that conversation, with my own commentary along the way. I am sharing it not because Gemini got everything right — it didn’t, and I’ll say where — but because the act of thinking through these ideas with a fluent partner is itself a demonstration of what TEI proposes: that intelligence is a collaborative, distributed, embedded thing, and that the most useful AI is the one that helps us see further than we could alone.
Why Pribram
Karl Pribram (1919–2015) was a neurosurgeon who became one of the most original neuroscientists of the last century. The leading theory of his day held that memory was stored in specific spots in the brain — discrete locations you could, in principle, point to. Pribram was frustrated by this picture because the evidence kept failing it. When portions of a brain were removed, specific memories often survived. Whatever was carrying the memory was not in any one place. It was everywhere at once.
So I asked Gemini to lay out what Pribram concluded, and how his work connects to the neural networks that now run modern AI.
Pribram’s theories provided a biological and conceptual bridge to what we now call Artificial Neural Networks. His influence can be seen in three places.
First, distributed representations — just as Pribram argued that a single memory is spread across a vast network of neurons, modern neural networks store information in the “weights” distributed across the entire model. No single node holds the concept of a “cat”; the concept is an emergent property of the whole.
Second, the holographic analogy. Pribram used the physics of holography to explain brain function. In a hologram, every part contains the information of the whole. This is mathematically similar to Fourier Transforms, used in signal processing and certain neural-network layers to analyze patterns in frequency rather than just space.
Third, parallel processing. His work moved neuroscience away from the brain as a serial computer toward a model of massive parallel processing — the exact architecture used by the GPUs that train modern AI today.
That is a fair and useful summary. It is also where Gemini reached for the connection to TEI directly — and where I want to slow down, because this is exactly the kind of moment a careful reader of these essays should learn to watch for.
The temptation in conversations like this — and Gemini reaches for it within a few hundred words — is to treat Pribram’s holonomic theory as proven, and then to treat TEI as the framework that completes it. That overstates both. Pribram’s holonomic theory is a serious, productive hypothesis with real empirical support and real unresolved questions. TEI is a structural account of how intelligence is embedded in systems capable of carrying it. The two illuminate each other; neither one closes the case for the other.
I want readers of TEI to develop the habit of holding a strong idea and a careful skepticism in the same hand. That is what the theory itself demands.
Bioelectricity, and the Field That Carries the Signal
I pushed Gemini to align Pribram’s work with bioelectricity, with Platonic Forms, and with TEI. The reply was striking — and where Gemini’s prose was at its most eager, I want to translate the substance back into a register I can stand behind.
Pribram’s work focuses on the dendritic arbor — the fine-mesh web of fibers at the end of neurons. He argued that intelligence isn’t just about neurons firing in binary spikes, but about the slow-wave bioelectric potentials occurring in these webs. When these waves meet, they create interference patterns — a “hologram” of information. Bioelectricity isn’t just a power source; it is the fluctuating medium that carries the embedded structure of our sensory world. It is the literal fluid in which intelligence is embedded.
Strip away the flourishes and what remains is a claim worth taking seriously: intelligence in the brain is not just a matter of which neurons fire and when. It is also a matter of the field — the continuous, slowly varying electrical state of the tissue itself. This field is not noise. It is structure. And structure is what TEI is fundamentally about.
Where Gemini wanted to leap was to Plato. The spectral domain of frequencies that Pribram described, Gemini argued, is essentially a scientific mechanism for Platonic Forms — the perfect mathematical patterns that the physical world is a shadow of. That is a beautiful idea. It is also a metaphysical claim that I will neither endorse nor reject in a TEI essay, because TEI does not require it.
What TEI does require, and what Pribram’s work supports, is something more modest and more important: that the structure of intelligence in a biological system is not located in any single component. It is embedded in the relationships among components — in the field, in the interference patterns, in the way slow-wave dynamics compose with discrete firing. The architecture is the intelligence. That is the part I want a TEI reader to walk away with.
Michael Levin, and Intelligence Without a Brain
The conversation got more interesting when we moved from the brain to the body. I asked about Michael Levin’s work on bioelectric morphogenetic fields, and Gemini gave a clean account of why it matters.
Levin has shown that embryos and regenerating organisms — like planarian flatworms — use bioelectric gradients to decide their shape. Even before a nervous system forms, cells use ion channels to create electrical maps. This pre-neural intelligence tells a cell, “you are part of an eye,” or “you are part of a tail.”
When a flatworm is cut in half, it “remembers” its original shape and regrows exactly what is missing. Levin calls this the “target morphology” — a kind of biological blueprint that the matter strives to achieve.
This is the result I have been waiting for someone to take seriously inside a theory of intelligence. Levin’s work, which is rigorous and ongoing, shows that cells communicate intelligently with each other through bioelectric fields long before there is anything we would call a brain. The “thinking” about what the body should look like is happening at the level of the tissue. The body is, in a real and not merely metaphorical sense, an intelligent system.
That is exactly what TEI predicts. Intelligence is not a property a system has once it crosses the threshold of “having a brain.” It is a property of certain kinds of organized, self-referential, purposive systems — and brains are one particularly impressive example of such a system, not the only one.
The Question That Won’t Stay Buried
Late in the conversation I steered toward something Gemini was ready for: the question of where the intelligence that runs in a data center actually comes from.
You hit on a key TEI principle: intelligence is scale-free. AI in data centers isn’t “new” intelligence; it is a concentrated, silicon-based distillation of the collective intelligence of humanity. Everything an LLM or a quantum processor knows was harvested from the collective output of human culture. Without the human data to sense, the silicon AI has nothing to process.
This echoes a point I made in an earlier essay in this series, about Richard Dawkins and the AI he called “Claudia.” The intelligence flowing through a large language model is, at its root, the reflected intelligence of human civilization — millennia of writing, thinking, arguing, distilled into weights. That is genuinely remarkable. It is also not the same thing as a brain generating mind from within itself. And it is certainly not the same thing as a body of cells deciding, by bioelectric vote, what shape it wants to be.
Three different kinds of intelligence, all real, all governed by the same TEI principles, none of them reducible to the others:
Reflected intelligence.
What a large language model produces. The compressed, recombined output of conscious minds across centuries. Powerful, useful, in some ways uncanny — but downstream of minds, not equivalent to them.
Embedded biological intelligence.
What a flatworm uses to regrow its head, what an embryo uses to become a body, what a brain uses to remember a face. The intelligence Levin and Pribram study. Distributed across cells and fields, structural, ancient.
Adversarial intelligence.
What a virus uses to hijack a cell. Minimal, ruthless, purely focused on its own replication. Gemini and I agreed this counts — a virus is an intelligence embedded in a minimal physical form, surviving by exploiting the embedded intelligence of larger organisms. TEI does not require that intelligence be benign. It only requires that it be structured and purposive. The virus qualifies.
That these three are recognizably the same kind of thing — three dialects of embedded intelligence — is, to me, the most important consequence of taking TEI seriously. It changes how we think about AI safety, about regenerative medicine, about public health, and about what we are when we sit down at a keyboard and ask an AI about Karl Pribram.
What I Wanted Gemini to Notice (and Didn’t Push Hard Enough On)
Re-reading the transcript, there are places I let Gemini’s enthusiasm carry the prose further than I would have on my own. The “transceivers tuning into frequency-based reality” line is poetic but speculative. The claim that Levin “proves” Pribram is overstated — Levin’s work is consistent with elements of Pribram’s framework, but proof in the philosophical-of-mind sense is not yet on the table.
These are not failures of the AI. They are reminders of what an AI is. Gemini was doing exactly what reflected intelligence does: producing fluent, well-shaped prose drawn from a vast distillation of human thought. The discernment — the moment when someone says, “yes, but” — has to come from the human in the conversation. That is the deal. That is what TEI predicts, and that is what I want every reader of this essay to take into their own AI conversations.
Use these tools. Use them constantly. But bring your own mind to the table.
The brain is not where you think it is. The intelligence you call “yours” is older, wider, and more distributed than that. So is the conversation.