The Brain's Best Computer Has No Inner Life

The conversation begins with a film more than fifty years old. Penrose has said elsewhere that *2001: A Space Odyssey* is his favorite movie, and he still thinks it is the science-fiction film that got the details right. People praise *Interstellar* as the scientific one; he does not think it is a patch on *2001*, which sweat the free-fall and a hundred other particulars. The opening bone-to-spacecraft cut still amazes him. The monolith he treats as an Arthur C. Clarke plot device — he was always a Clarke fan — not as a theory of innovation.

What sharpens the talk is HAL 9000, which decides to remove the astronauts because it believes they will interfere with the mission. Fridman asks whether HAL did the right thing, and whether a future system should. Penrose hears a prior question: are we supposed to believe HAL was actually conscious? The film plays it that way. HAL shows something like pain and an inkling of what it means to die, and it does not like the idea of being turned off. If HAL is only a very sophisticated computer without those feelings, turning it off is a technical act. If it is conscious, turning it off may be wrong.

That fork, he says, is the unspoken premise of a lot of AI research: pile up computation past some threshold and consciousness will come along. Then you have problems. One reason people want AI is to send a device to a distant planet without paying to bring a human home. If the device is a conscious entity, leaving it there is immoral, and thinking that building an AI removes your responsibility to it is incorrect. He once tried to paint Douglas Hofstadter into that corner — certain large integers would have to be conscious — as a reductio. Hofstadter leapt into the corner and found it perfectly reasonable. People think this, Penrose says, basically because they cannot think of anything else.

Most of what we actually call computation in the head, Penrose thinks, is done by the cerebellum. He comes to the subject from outside, and some anatomical facts strike him that people hardly mention. The homunculi are a mess: vision at the back, feet at the top, right hand on the left side, sensory and motor maps facing each other rather than sitting adjacent. It is about the worst organization you can imagine. It cannot just be a mistake in nature. Something else is going on. It does not look like a circuit diagram.

When he first thought about these things he was told the cerebellum had maybe half as many neurons as the cerebrum, the convoluted thing on top that everyone talks about. Now they tell him it has far more neurons, more connections, much more going on computationally. It looks a bit like a ball of wool at the back underneath. A pianist does not consciously will the little finger onto the key at the exact millisecond; a tennis player does not calculate which muscle fires in which direction. Consciousness may be in the feeling of the music or the tactic of angling the ball. The precision is cerebellum. And as far as we know — slightly controversial — the cerebellum is entirely unconscious.

That is the hinge of the conversation. If computation produces consciousness, here is an incredible amount of computation that produces none. So what is the difference? The cerebrum's crossover and its odd lobes look completely inefficient from a computational perspective. There must be a lot of computation going on, but the cerebellum seems better at doing that. The idea that the brain is just a complicated computer, he says, is completely missing the point.

Fridman asks for Gödel's incompleteness theorem and how it touches computation and consciousness. The idea, Penrose says, formed in his first year as a graduate student in Cambridge. As an undergraduate in London he had worried with a colleague, Ian Percival, that Gödel seemed to say there were things in mathematics you could never prove. In Cambridge, while officially doing algebra and geometry, he sat in on three courses that had nothing to do with his work: Hermann Bondi on general relativity, Paul Dirac on quantum mechanics, and a course on mathematical logic. Bondi made relativity alive. Dirac was organized and never seemed excited. The logic course gave him Turing machines and Gödel, phrased in a way Douglas Hofstadter's book does not.

Suppose what you mean by proof is something a computer can check: a lot of steps, an algorithm that says whether you followed the rules. Why do you believe a "yes" is actually true? Because you have looked at the rules one by one and seen that they only yield truths. Gödel then constructs a mathematical statement of the very kind the system is supposed to judge. By the way it is built, and what it means, you can see that it is true — and that it is not provable by the rules you were given. Your belief that it is true depends only on your trust that those rules never lie, and on seeing that the coding was done correctly. That, Penrose says, blew his mind. It is as good as any proof, and it lets you transcend the system.

What enables you to see beyond it? Understanding of what the system is actually saying, and of what the constructed statement is saying. That quality is not governed by rules. It is not a computational procedure. People say one set of rules is as good as any other. That is not true. You have to understand what the rules mean, and why that understanding gives you something beyond the rules themselves. Understanding why the rules give you truth enables you to transcend the rules. Already then, the thought entered his mind that understanding — intelligence, even consciousness — sits outside the rules.

He has always concentrated on understanding rather than creativity. Someone can put funny marks on paper, a machine can do that too, and he does not know how to work with the claim that machines cannot be creative. Understanding is different. You can see, via Gödel, that it is something, even if it is very hard to put a finger on. He has never, as a mathematician, produced a formal definition. What he can do is point at its shape.

He sometimes draws a cartoon. In the foreground a mathematician is proving a theorem — there is a little joke in the theorem, never mind — and a saber-toothed tiger is about to eat him from the undergrowth. In the distance his cousins are growing crops, building shelters, domesticating animals. In front, a mammoth is falling into a pit and people are about to grab it. The quality of understanding is not the mathematician's private toy. It is what let ancestors live: seeing the trap, the path of the mammoth, when to close the net, before the action. Natural selection paid for that. Mathematics is a later branch.

Asked to dance around a definition, he offers something slippery: standing back and thinking about your own thought processes. It is the Gödel move. You are not following the rules; you are thinking about the rules. You are doing something, then you think, what the hell am I doing? You take a step back outside the game you have been playing. You are not playing anymore. You are thinking about what it is to play. It is not a precise description. It somehow feels very true. That reflection, he hopes, might someday be unearthed as what gives conscious beings their advantage, and he thinks it goes way back.

Fridman asks how far back, if understanding is evolutionary, the line goes. Penrose does not think it is limited to humans, and he does not think it starts with primates. He describes African hunting dogs, probably from a David Attenborough film, dividing into two groups. One hides by the river. The other rounds up the antelopes and drives them toward the water, where they slow and the waiting dogs pounce. They have planned this somehow. He has no idea how. There is an element of conscious planning, as far as he can see. It does not feel like a bottom-up neural net trained on a zillion examples, picking up signals you are not even aware of. That does not feel like understanding.

Elephants move him more. The leader of a troupe is a female. At a certain point she makes a long detour; the others follow. It is where her sister died. Bones are lying around. They pick them up, hand them round, caress them, put them back, and go. What the hell are they doing? There is no clear, direct connection with natural selection. There is some deep feeling going on, tied to their conscious experience. He tends to think this sort of thing is overall advantageous, but not as a module carved step by step for survival. It spills from the experience itself.

Octopuses show the same sort of quality people are willing to call consciousness, and their evolutionary route is completely different. Does it go way back to a common ancestor, or did it come separately? Either answer would matter for what consciousness is. Fridman wonders whether we have to throw away deterministic machines as we know them. Penrose's hope is that something is simple. He is not saying consciousness has to be biological. We do not know. We also do not know what material is.

He still thinks of himself as a materialist in the sense that nothing mystical blows in from who knows where. He does not like the word, because we do not know what material is. If understanding is not computational, it still has to sit on some physical process. Newtonian mechanics can go on a computer. Continuum versus discrete bits he came to think is probably not the catch; you can get close enough, though he might retract someday. General relativity, from Bondi's course, also goes on a computer. People now calculate black holes spiraling into each other and the shapes of the gravitational waves. LIGO sees those signals. That is a vindication of computation describing Einstein. Close. Is it close enough?

Then Dirac's first lecture. A particle, Dirac said, is usually here or there; in quantum mechanics it can be here and there at the same time. He got out a piece of chalk — some people say he broke it in two — to illustrate a superposition of locations. Penrose's mind wandered. He does not remember the explanation, something about energy that seemed to have nothing to do with anything. It is probably just as well. It would have been one of those remarks meant to calm you down. He has worried about it ever since. Somewhere the positions become one or the other, and that is not part of the Schrödinger equation.

Follow Schrödinger and you can put it on a computer, modulo the continuum. The theory is incomplete, and in a certain sense not quite right, because to relate to the world we see you have to break the equation: reduction of the state, collapse of the wave function. Schrödinger was appalled. The cat is not a celebration of quantum weirdness. It is a protest: look, my equation leads you here, a cat dead and alive at once; something is missing. Einstein, Schrödinger, and Dirac, read carefully, all wanted something beyond the theory we have. Dirac was cautious, but in the right place he says quantum mechanics is a provisional theory; we need something that explains the collapse. Penrose's view is that the missing piece is the same kind of gap the Turing machine has about consciousness, and that it comes in when gravity is involved.

The conflict, worked out much later, is between Dirac's superposition and Galileo's principle of equivalence, the seed of general relativity. In free fall you eliminate the gravitational field, at least locally. Galileo's fireworks explode into a sphere of sparks that stays a sphere as it falls, as though there were no gravity. Astronauts see the Earth right there and feel none of it; they are falling freely. Einstein built a theory on that. Superposition says a mass can be here and there at once. Equivalence says the gravitational field depends on where the mass is. A lump in two places at once has two gravitational fields. Which one is the world?

At tiny scales the clash does not matter. When the superposition is "heavy" enough, Penrose thinks nature forces a choice. That choice is the collapse, and it is not a computation. Whatever consciousness is, it is not a physical process that can be described by computation — but it nevertheless could be physical. People hear "quantum" and think the mystery is already in ordinary quantum biology. He wants something more outrageous: you have to go where conventional quantum mechanics goes wrong. Photosynthesis uses quantum mechanics in a basic way, beyond ordinary chemistry. Neurons might too. Following Schrödinger is still computable. Gravity, in his minority view, must change the structure of quantum mechanics, not merely sit inside it. "Quantum gravity" usually means the opposite: make gravity fit the quantum formalism. He wants a two-way street. He does not have the theory. That is the trouble.

He had meant to write a popular book in retirement. A radio discussion between Edward Fredkin and Marvin Minsky hurried him. They imagined walking into a room where two computers at the far end had already exchanged more ideas than the entire human race. He knew where they were coming from. He did not believe them. He wrote *The Emperor's New Mind* around the time Hawking was writing *A Brief History of Time*; Martin Gardner did the foreword after Hawking got Carl Sagan. Writing it, Penrose realized he knew nothing about neurophysiology. He read up, looking for how nerve signals could preserve quantum coherence. Everything he found said the electrical traffic through wet, ionic tissue had no chance of isolation. Hopeless. He more or less gave up at the end of the book.

Then Stuart Hameroff wrote. You do not know about microtubules, do you? They are the little tubes that pull chromosomes apart in mitosis, present in pretty well every cell except red blood cells. Hameroff's day job is general anesthesia: turning consciousness off reversibly so the patient feels no pain, then turning it back on. The gases have varied chemistry and an unclear mechanism. Hameroff thinks they act on microtubules. Penrose was excited by the symmetry of the tubes. Symmetrical structures, he had reason to believe, preserve quantum coherence far better than messy ones; once the state leaks into the environment, you have lost. He did not yet know the difference between lattice types. The claim is more than "it's quantum mechanical." It is that the measurement process — state reduction — happens there, and that is where non-computability comes in.

Orchestrated objective reduction, Orch-OR, splits in two. OR says collapse is an objective gravitational event, each one an element of proto-consciousness, the building block of experience. Orch says the brain somehow orchestrates those events into a genuine conscious life. Penrose is blunt about the second half. He happily put the word "orchestrated" down. What it really means is perhaps even more unclear than the word "material." Experiments are not nothing. Dirk Bouwmeester, splitting time between Leiden and Santa Barbara, has been developing a test of whether quantum mechanics without gravitational effects must be modified. Hameroff's group looks at whether anesthetic gases directly affect microtubules — one of the few handles we have, Penrose says, because they specifically turn consciousness off. The evidence, he will not overclaim, points in that direction. The view is a minority among minorities, but it is no longer just them and a few friends.

Penrose sorts physics theories into superb, useful, and tentative. Fridman asks whether we will ever have a superb theory of intelligence and consciousness. We might. We are a long way from it. He does not think we have even entered the tentative scale. There is no picture a majority accepts. Most views are computational in one form or another, and even then people are unclear: they say consciousness is computational, then talk as if it is not, and cannot say what it is computing. Early quantum physicists often ran the dependence the other way: collapse happens when a conscious being looks. A few distinguished physicists still hold a version of that. Penrose's view is almost the exact opposite. Nature chooses, all the time, in some non-computational way we do not understand. That choice is the building block. Consciousness depends on the state becoming one or the other, not the other way around.

He tends to sit understanding in the middle of a three: awareness, understanding, intelligence. Could you call an entity intelligent if it does not understand? Could you say it understands unless it is aware? He is using terms he cannot define. They are somewhat poetic, not mathematical. As a mathematician he can at least point to the connections, and understanding is where Gödel lets him say something. What it is to perceive the color blue, or whether your blue is his, or synesthesia mixing colors and sounds — from the fundamental physics, we are way off. The Orch part, how the brain binds proto-conscious events into a life, might be beautifully simple. At the moment we are completely in the dark.

Fridman turns to the 2010 book *Cycles of Time* and a universe before the Big Bang. Penrose does not like calling it another universe; these eons are not separate. He grew up with the steady-state picture — no bang, hydrogen created as space expanded — which he found philosophically nice and which the microwave background killed. Fred Hoyle used "Big Bang" as an insult; the name stuck, as with Schrödinger's cat. The crackle on a television is already that background, from about 380,000 years after the start. Einstein had wanted a universe that was always there. Friedmann and Lemaître, solving his equations, found a concentrated initial state Lemaître called the primordial atom.

The equations work beautifully, he says, except for inflation: in a fraction of a second whose denominator has something like 32 to 36 digits, the universe is supposed to have expanded exponentially for no apparent reason, via an invented inflaton field. He thought it completely crazy. The deeper issue is the second law. Things get more random as time goes on, so going back they get less random. The earliest thing you can directly see is the microwave background, and its most striking feature is that it is random — a Planck spectrum, a thermal curve, entropy at a maximum. Then it has been going up ever since. He calls that the mammoth in the room. Why don't cosmologists worry?

They are looking at matter and radiation in a maximum-entropy state. What they are not seeing directly is gravitation. Gravity was very, very low entropy, low by its uniformity, which you also see in the microwave sky. That, to him, is the biggest problem in cosmology. Nobody seems to worry. People say they have solved all the problems. They think inflation solves it. It cannot. Inflation can stretch uniformity that is already there; it cannot manufacture it from a mess. That, he says, is completely clear, and the field goes around it.

He started from boredom. Twenty-first-century supernovae showed the expansion is exponential; Einstein's cosmological constant, introduced for a static universe and later called his greatest blunder, was not a blunder. The future gets more and more boring. Black holes gulp clusters of galaxies. Andromeda's hole swallows ours. Then you wait an unbelievable time and Hawking evaporation pops the holes away. Colder and colder. Who is going to be bored? Not us. Photons, and photons do not experience time the way we do. In the way Penrose used to treat gravitational radiation, infinity is just another place. You can squash it down if there is no mass. Photons get there. The usual view says there is nothing on the other side — just a boundary.

He reaches for Escher's circle-limit angels and devils, which Escher got via Coxeter. The creatures crowd toward the edge, same shape, smaller and smaller. Their infinity is a circle to us. Hyperbolic geometry makes that precise. Photons have no mass. Our clocks are precise because of two equations: E=mc² and E=hν. Mass is frequency; a massive particle is a clock. No mass, no clocks, no rulers, no scale — only conformal structure, the shapes the angels keep as they shrink. In the remote future, after holes evaporate, that is what is left. At the Big Bang, the opposite: hotter, denser, particles near light speed, mass negligible against energy, scale gone again. Squash one conformally and you get the other. Hot and dense, cold and rarefied, look different, feel different, are almost the same.

The remote future of a previous eon is our Big Bang. Signals can come through. The two main ones he names involve black holes: collisions that dump gravitational waves into the next eon. He and a colleague, Vahe Gurzadyan, have a paper on this, including a pass at the Fermi paradox. If civilizations can send something through the boundary, why haven't we heard? Because we do not know how to look. SETI is still reasonable. Another place to look is not the neighbors who got here first, but civilizations so successful they harnessed signals for whoever could pick them up on the other side. Cosmologists find the idea wild. To him, photons needing somewhere to go is not such a wild idea.

Asked for the most beautiful idea in physics or mathematics, Penrose does not say gravity or quantum theory. He says complex analysis. Integers, then fractions, then the reals for a continuous line. Then x² = -1 has no real solution, so you put in i. One little addition and structures you never imagined fall out. The construction then turns out to be the crucial basis of quantum mechanics; the i in Schrödinger's equation is not decoration. On whether mathematics is invented or discovered, he does not hedge. Absolutely discovered. More like archaeology than you might think. That fits the rest of his life: a structure already there, formal systems that cannot exhaust it, a mind that can stand outside the system and see.

The meaning of life he will not call a stupid question, though many scientist colleagues do: molecules came together, so what? He thinks there is more to it. What that more is, he does not really have an idea. It might be tied to the mechanisms of consciousness. The mysteries are tangled: consciousness, quantum mechanics, how the quantum fits the classical world, even the mystery of complex numbers — mathematical puzzles that seem to bear on how the physical world operates. "We're scratching the surface. We have a long, huge way to go before we really understand that." Fridman closes with a passage from *The Emperor's New Mind*: children are not afraid to pose basic questions that embarrass adults, and the obviousness that a conscious mind cannot work like a computer is the kind of obviousness a child can see.