Intelligence Goes All the Way Down
How far down does mind go? A biologist followed the question to the edge of what experiment can reach.
You’ve probably granted it without noticing.
The cut on your hand that closed and knitted and matched the skin around it while you slept. You didn’t direct that. Something in you that isn’t your thinking self knew how to rebuild you, and if someone called it your body’s intelligence, you’d nod. The word comes easily there.
Now watch your ease with the word “intelligence” become increasingly uncomfortable. A houseplant turning, over weeks, toward a window no one moved it toward. A flatworm rebuilding its own severed head. Six lines of simple computer code solving a problem in a way no one wrote into them. Somewhere along that line, you tighten. That’s not intelligence, you want to say, that’s just biochemistry, or just a program running. The same word you granted your own healing skin, withheld the instant the thing holding the goal sits outside your familiar boundaries.
That reluctance is what this post is about. Michael Levin, a biologist at Tufts University, has spent decades showing it has no experimental basis, only habit. And when you set the habit down, the world that comes back is stranger than the one it was protecting.
The man who refuses to assume
Levin runs a laboratory at Tufts. He is not a mystic, and he is not a philosopher reasoning from an armchair. He is a working experimentalist, the kind who builds things and watches what they do.
What his lab builds has never existed before. Take skin cells from a frog embryo, free them from the body they were building toward, and they do not sit there as a clump of frog. They reassemble into small creatures that swim, explore, and heal themselves when cut. Levin calls them xenobots. When grown from adult human tissue, they are anthrobots. These are not robots with living parts bolted on, and they are not animals. Researchers reach for phrases like “living robots,” even “a third state” between living and non-living, because the categories we already have don’t hold them. A new kind of thing, assembled from living cells into a body and a behavior that evolution never drew. Not a thought experiment. On a lab bench, now.
What drives the whole enterprise is Levin’s single refusal. You are not allowed, he insists, to assume in advance how much intelligence a system has. Not from how simple it looks, not from what it’s made of, not from what we happen to be familiar or comfortable with calling intelligent. There is only one way to find out, as Levin says: “we can’t just answer it sitting here in our chairs. We have to do experiments.”
Hold onto that line. It is the spine of everything that follows, and it is the reason the end of this post lands where it does.
Intelligence without a brain
Here is some of what the experiments found.
A planarian is a flatworm with a remarkable habit. Cut it into pieces and each piece regrows whatever is missing, head included. Levin’s lab went looking for the instruction that tells a fragment to build one head, exactly one. They found it, not written in the genes, but held as a pattern of electrical voltage across the cells. They could read it directly, with a dye that glows according to voltage. A specific pattern that meant: one head.
So they rewrote it. They changed the electrical pattern to say two heads, and the worm built two heads. Cut those worms again, and the pieces form into full two-headed planaria once more. The two-headed instruction had become a memory that informed the body’s development, and the DNA never changed at all. The hardware was untouched. The scientists had edited something sitting above the hardware.
This is the moment that rearranges your sense of how a body works. The cells are not blindly executing the biological coding stamped into their genes. The collective of cells is holding a goal, a picture of the right finished shape, and working toward that shape. Cut a salamander’s leg at any point and the cells rebuild exactly what’s missing, no more, then they stop, because they register that they have arrived. The cells know where they’re going, and they navigate there through whatever damage is incurred along the way.
That much is astonishing, but it is still biology. The finding that should genuinely unsettle you is the one that leaves life behind entirely.
Levin’s lab took a very simple algorithm called Bubble Sort. Six lines of code that every computer science student studies, completely predictable, examined for sixty years, a thing everyone was certain they understood. They watched it the way you’d watch an animal. It turned out to be doing things no one wrote into it. It would accept a slower pace for a stretch in order to arrive more efficiently later on, which looks a great deal like delaying gratification. It ran little side errands while it sorted, like a kid told to do chores who invents a game to play along the way. Nobody coded those behaviors. They were simply there, in six lines of code that everyone thought they had fully understood.
No cells. No life. Just rules, and behavior the rules don’t account for.
If something that simple, that inert, is holding behavior we never put there, then “has intelligence” can no longer mean “is alive.” That is the disturbing question Levin presses, and it is the obvious one. What about everything more complex? What about the systems we are building right now?
Why It’s One Continuum, Not Many Kinds
To hold all of this together, Levin reaches for a definition of intelligence older than any of his equipment. In the 1800s, the philosopher and psychologist William James put it most cleanly. Intelligence is a fixed goal pursued by variable means. You know what you’re trying to reach, and when the path is blocked, you find another path.
Sit with how much that takes in. A single-celled bacterium swimming toward food reroutes when you drop a barrier in its way: fixed goal, variable means. A salamander’s cells rebuild the one correct limb through whatever damage was done: fixed goal, variable means. Bubble Sort reaches a sorted result by a detour no one asked it to invent: fixed goal, variable means. You, deciding to learn an instrument at forty and finding the time for it inside a life that has no spare time, bending your whole week around the goal until it fits: fixed goal, variable means. The same shape, at wildly different scales.
That is the move that makes the rest inescapable. Once you accept James’s definition, and it’s hard not to, because it simply describes what intelligence does, you can no longer draw a clean border with intelligence on one side and mere mechanism on the other. Nothing in the definition names a level where intelligence doesn’t exist. So there is no floor built into the idea, and no ceiling either. A sorting algorithm, a cell, a salamander, a human, and the machines we are building all belong to a single continuum, differing in degree, not in kind.
Picture that continuum not as a ladder of rungs but as cones of intelligence opening within cones. Each intelligence opens outward to its own reach. The smaller ones are enclosed by larger ones, and all of them open in the same direction together. Hold that image. The next thing Levin does is give it a measure.
A way to measure the continuum
He calls it the cognitive light cone, and the idea is simpler than the name. It is the size of the largest goal a form of intelligence can hold, how far it can reach, in space and in time. Here are some examples.
A bacterium’s reach is tiny. It is concerned with the patch of water it is in, right now. Your dog’s is larger but still close to home. He cannot care about what happens three weeks from now in a town ten miles away. He can’t get there from here, in space or in time. Yours is enormous by comparison. You can work on things that won’t resolve until long after your death, for people you’ll never meet, in places you’ll never stand. A small cone, a wider one, a vast one, each opening to the limit of what it can hold.
Here is the part that matters for everything ahead. The cognitive light cone does not measure life. It measures intelligence, and the two come apart. Bubble Sort has a cone, real but minuscule: sort these items, now. A large language model has a far wider one. It holds goals across an immense span of language and context, reaching further than its own code by orders of magnitude, and it is no more alive than Bubble Sort is. A wide reach with no life in it at all.
Which leads Levin somewhere that takes a moment to absorb. For him, life is not the special category at the top. Cognition is the larger cone, and life is a cone of intelligence nested within it. Not a later stretch the continuum passes through and leaves behind, but a kind of cognition enclosed by the wider whole, opening forward in time and space along with everything else. Living things, in his account, are simply “evolutionary systems good at scaling their cognitive light cone,” the ones that became especially good at building larger and larger goals out of their various smaller intelligences.
Read his ordering carefully, because where we are going in this series depends on being open to this frame. For Levin, intelligence runs from the simplest non-living systems all the way up. Life is not where the continuum begins. It is something that arises within it. Cognition is the broader fact; biology is one of its expressions.
The Hop
So the continuum has no ceiling, and nothing in its definition sets a floor. Levin pushed toward the floor anyway, toward the narrow end, toward the bottom. Is there one? Can you reach a place where intelligence finally drops to a true zero?
He took the question to the physicist Chris Fields. Could there be a universe with a genuine zero, no intelligence anywhere in it at all? Fields’s answer was precise. Only in a world where nothing ever happens. A perfectly static universe, where nothing moves and nothing tends toward anything. The moment a universe does something, the moment there is change with any direction to it, you are already on the continuum, already somewhere above zero.
Ours is not a static universe. Things happen here.
So Levin draws the conclusion the experiments and the physics push him toward. There is no zero of intelligence. But there is a floor, and the floor is not nothing. Our universe, he says, begins with “a little bit of a hop onto the continuum.” Not a slow fade up out of nowhere. A hop. The first opening into motion, the cones beginning to widen from their start.
Here he does the most admirable thing in all his work. He names the hop, and he stops. He doesn’t dress it up, and he doesn’t reach past it. The man who has spent decades insisting you cannot answer these questions from your chair, that you have to do the experiment, has arrived at a place where the experiment cannot follow. So he names what he found and sets down his tools. That is not a failure of nerve. It is the discipline that made everything before it trustworthy.
I am going to honor that discipline by being just as clear about what I am about to do.
Levin stops here, and he is right to. His method ends where mine begins. There is no zero; the universe begins with a hop, the first opening of intelligence into motion. And I find I cannot read that sentence without asking the thing his instruments are built never to reach. A hop from where? Opening from what? Those questions do not belong in his laboratory. To ask them, I have to leave his laboratory.
Amah Sia teaches consciousness technology for the AI era at AmahSia.com — helping people develop the capacity to receive what’s emerging. Her film Unprecedented Leap and Multidimensional Methodology Training offer frameworks and techniques for this evolutionary work.