400 Spins in a Pocket, and a Universe Too Small to Store Them

Lex Fridman introduces Leonard Susskind as a Stanford theoretical physicist, founding director of the Stanford Institute for Theoretical Physics, and one of the people widely called a father of string theory. The first question is Richard Feynman: they worked together and were friends, and Fridman wants to know how that changed him. What Susskind saw was someone who could do physics in a deeply intuitive way. Feynman's style was almost to close his eyes, visualize the phenomenon, and through that visualization outflank highly mathematical, sophisticated technical arguments.

That way of working, Susskind says, was already natural to him. Feynman did not rewrite his mind; he validated it. Here was someone successful who did physics more simply, more directly, more intuitively, which made Susskind look at his own habits and conclude that you can get away with this in practice. Asked whether intuition and visualization come first when he thinks about quantum mechanics, black holes, or string theory, he does not hesitate. He tends not to think about the equations or the symbols. He tries to visualize the phenomena themselves. When an insight seems valid, he might convert it to mathematics.

He is careful about the conversion step. He is not a mathematician, not a natural one: good enough, he says, but not great. The sequence is intuition, visualization, maybe a few scribbled equations, then an attempt to put it into math — and his experience is that other people are better at that translation than he is. Coming from someone whose public reputation is tied to a highly mathematical subject, the order of operations is the point. The math is downstream of a picture.

Fridman presses the obvious tension: Susskind has spent a career on ideas that are deeply counterintuitive, so how does anyone visualize them? Susskind takes the word apart. Intuitive, in the everyday sense, means the ability to think with the classical physics we evolved with — throwing stones, splashing water. Quantum physics, general relativity, and quantum field theory are unintuitive in that way. Familiarity grows a different intuition. He has always called it rewiring. He and many of his friends can now think more easily quantum-mechanically than classically. They have gotten used to it.

That rewiring has a ceiling. Being able to think somewhat quantum-mechanically does not mean you can think like an electron. A cleaner example is extra dimensions. He believes we are fundamentally wired to visualize three dimensions. He cannot even visualize two dimensions or one without embedding them in three: a line is a line in space, or a line on paper, and the paper sits in three dimensions. He has never been able, in some abstract pure way, to hold one, two, four, or five dimensions in his head, and he does not think that will ever happen. The neural wiring is set up for three.

People do learn ways to think about five, six, seven dimensions — mathematical ways, visualization tricks — but those are different from the native three-dimensional picture, and he doubts we can ever become completely comfortable. Fridman romanticizes creatures in a two-dimensional space imagining 3D. Susskind's answer is probably the way we imagine four: mathematics, equations, tricks. The recurring posture of the conversation is already here. Admit the wiring. Then go around it with tools.

Feynman had a bit of an ego. Fridman asks whether ego in science is powerful or dangerous. Both, Susskind says. You have to have the arrogance to say: nature is very, very hard, and I am smart enough, I can win this battle. You also have to have the humility to know you are very likely wrong on any given occasion. Everything you are thinking could suddenly change. Young people can arrive speaking a language you do not understand, and you will be lost and flabbergasted. You had better recognize that you are very limited, and you had better be able to tell yourself you are not so limited that you cannot win. It takes a special kind of person to hold both.

Asked whether he ever looked at himself and wondered if he was completely wrong — about the whole thing, meaning his ability — he draws a distinction. Those were not the doubts he had. He came from a working-class background and was uncomfortable in academia for a long time. The discomfort was not about talent. It was the unease of being in an environment his family had never participated in. He did not learn there was such a thing as physics until he was almost 20. He does not remember worrying about whether he would get a job. That thought had not occurred to him.

Fridman, also uncomfortable in academia, offers a music metaphor: Susskind plays a different kind of music. Susskind says he did not think it so much as feel it. Thinking is one thing; feeling is another. He felt like an outsider until a certain age, when he suddenly found himself the ultimate insider in academic physics. The transition was sharp, and he was not young — probably 50. Fridman calls it a phase transition, never quite in the middle. Susskind agrees. His way of thinking was different, his approach to mathematics was different, his social background was different; these days half the young people he meets have professor parents. Then, at some point, he was at the center of a certain kind of physics, not the only one, and all of that went away in a flash. How he develops ideas now is both solitary and social: more time alone when younger, and at Stanford a daily practice of brainstorming with ex-students and colleagues, less time staring at a blank page.

Fridman's question about hopes for quantum computers arrives dressed as "machines that leverage quantum-mechanical ideas." Susskind corrects it. You can simulate a quantum system on a classical computer by solving the Schrödinger equation. The classical machine is made of quantum mechanics, as everything is, but it is not functioning as a quantum system. It is solving equations. A quantum computer is truly a quantum system doing the things you programmed. Program a quantum field theory classically and the machine is not running as a quantum field theory. Program it on a quantum computer and the device is carrying out quantum operations, limited by the uncertainty principle, measurable at the end. You are building a real version of the system.

That is why classical computers are, in his word, enormously limited here. Four hundred spins are not many; 400 pennies fit in a pocket. Storing the quantum state of 400 qubits would take more information than can possibly be stored in the entire universe, packed so tightly that you could not pack any more. A quantum computer made of 400 qubits, by contrast, can do everything 400 qubits can do. The contrast is not about clock speed. It is an information bound.

A handful of algorithms are known, as a mathematical statement not yet exhibited in the laboratory, to have exponentially more power than classical machines. It is also increasingly clear that the number of such problems is very limited. Only very special cases show that much advantage. Factoring large numbers, the famous application, does not seem to have much to do with quantum mechanics. That a quantum computer can factor quickly looks to him almost like a fluke: extremely special, rare, and not obviously a growing class. The real inventory is quantum systems themselves — chemistry, solid-state physics, materials, quantum gravity, quantum field theories, some of them already applied sciences.

When pencil-and-paper and classical computers run out, you build a version of the system, run it under controlled circumstances, change it, manipulate it, and measure. Why not just experiment on the real system? Real systems are limited. You cannot change them, cannot slow them down to poke into them, cannot modify them arbitrarily to see what happens if you nudge a parameter. The value is a manipulable slice of a quantum universe, not a quicker laptop.

Fridman asks whether any of this reaches the middle scale — brains, biological systems about a meter across. Among the people Susskind actually talks to, the excitement is black holes: big objects, many degrees of freedom. There is another big quantum system, a large quantum computer, and one thing they have learned is that the physics of large quantum computers is in some ways similar to the physics of large quantum black holes. That is the relationship they are using. Brains are different. Black holes function in a very quantum-mechanical way. He is not a neuroscientist and does not even have many neuroscientist friends, though he would like more. Among the few he has asked, the conviction is that the brain functions classically: it does not intrinsically use entanglement or coherent superposition.

Are they right? He does not know. He sort of hopes they are wrong, because he likes the romantic idea that the brain is a quantum system. He thinks probably not. Macroscopic and quantum are not opposites in any case. Materials can be large and still be made of quantum parts. Topological insulators and, more bluntly, superconductors are chunks of stuff whose properties depend very strongly on quantum mechanics; you cannot analyze them without quantum tools. Asked what you gain by treating the universe as an information-processing system, he is dryer. All systems are information-processing systems. You poke them, they change a little, they evolve. There is no extra magic to humans.

Consciousness and intelligence sure feel like magic, Fridman says. Susskind agrees that they do, then puts a fence around his own authority: if you ask him about physics, his thoughts may be worth something; if you ask him this, he is not sure they are. Introspection, he is pretty sure, gets the mind wrong. Everything he has heard about how the brain functions is counterintuitive. There are neurons that detect vertical lines and different neurons for lines at 45 degrees. He never imagined whole circuits devoted to vertical lines. Holding a finger vertical, horizontal, or at an angle feels like the same circuits. Compartmentalization is not what psychological introspection would have guessed. We will not get it right that way.

How will we? He thinks computer scientists might, eventually. He does not think anyone is near it, or even thinking about it in the right way. Eventually we will build machines complicated enough, partly engineered and partly evolved — maybe evolved by machine learning — and we may start to discover mechanisms with implications for how we think and for what this consciousness thing is. Then we can do experiments on them and answer questions introspection cannot touch.

Fridman notes that even string theory looked monstrous and then peeled down, and asks whether that method might reach biological systems. Physics always begins by finding the simplest version of something. Superconductors look like a monster of electrical and magnetic properties; boiled down, they are a simple quantum phenomenon called spontaneous symmetry breaking, already familiar from other contexts. Yes, they take complicated things and make them simple. What they do not want to do is take things that are intrinsically complicated and fool themselves into thinking they can make them simple. He does not want them simpler than they really are. Whether the brain ultimately runs on simple rules or is just complicated, nobody really knows.

Fridman sketches a physicists' chain about AI: intelligence is information processing, so there is no theoretical reason we cannot build human-level intelligence, and once we have, the same method should yield superhuman intelligence easily, exponentially. Susskind wishes he knew what to make of that chain, for a particular reason. He has a second job: consulting for Google, for Google X, as senior academic advisor to a group of machine-learning physicists. It sounds crazy, he says, because he knows almost nothing about the subject. On the other hand, he is good at giving advice, so he gives it.

He watches young physicists enter a real problem: why machine learning works as well as it does, why it generalizes. Nobody really seems to understand. He sorts the people who think about it into camps. Engineers are incredibly smart and tend to think less about why than about how to use the thing; they are also the ones who produced the data showing it works much better than you have any right to expect. The mathematics of tensor networks used to describe quantum systems is not all that different, structurally, from the networks used in machine learning. More good young physicists are being drawn in. They are super smart, he says, and not so arrogant about their physics backgrounds that they think they can do what nobody else can. The physics way of thinking will add value, and he thinks it already has. Timescales for breakthroughs, from his long experience of being surprised, can be a day or 20 years. Quantum gravity he once put at 500 years; it turned out closer to 20 or 25 from the moment he thought it was 500.

What is the dream of string theory? Susskind does not think it is different from the dream of fundamental theoretical physics: a unified understanding. He does not like thinking of string theory as a subject unto itself, with people called string theorists practicing a thing called string theory. He prefers theoretical physicists trying to answer deep questions about nature, especially gravity and its connection with quantum mechanics, who at present find string theory a useful tool. He does not like being referred to as a string theorist.

He was there at the invention. The first life, late 1960s and early 1970s, was hadrons: protons, neutrons, mesons. Experiment made it clear those particles could vibrate, rotate, and do what a little closed string can do. It was and is a valid, experimentally tested theory of hadrons — a done deal. It had a second life as a theory of gravity: the same mathematics on a scale 19 orders of magnitude smaller than a proton. The value, as he tells it, is not a slogan about unifying everything. In many ways the constructions are mathematically rigorous enough that you can find structures containing both quantum mechanics and gravity, and you can test ideas for internal consistency even if the interesting objects are 19 orders of magnitude too small for a laboratory.

Thirty-five or forty years ago, people very seriously questioned whether gravity and quantum mechanics were consistent. Stephen Hawking was famous for that, rightly so. Now nobody questions the consistency anymore, because there are mathematically precise string theories that contain both in a consistent way. That certainty — that they can coexist — is not a small thing. It is a huge thing. Would Einstein be proud? He might be appalled; Susskind does not know. He would certainly be struck. At this time, Susskind thinks that may be string theory's biggest contribution to physics: illustrating almost definitively that quantum mechanics and gravity are closely related and not inconsistent.

Could something deeper still sit under string theory? You could ask the same of quantum mechanics, Susskind says. He treats string theory as an example of a quantum-mechanical system that contains gravitation. Is there something under quantum mechanics, perhaps deterministic? He hands the question to his friend Gerard 't Hooft — a name he jokes is hard to say and easy to spell, the only person he knows whose name begins with an apostrophe, a hero of his, a little younger, not as famous as he should be. 't Hooft believes the world has a classical, deterministic structure which somehow, by a mechanism he has a hard time spelling out, emerges as quantum mechanics: not just the wavefunction, but uncertainty, entanglement, the whole package.

Susskind's own stance is the conversation's cleanest methodological line. This is where you have to be humble. He does not think anybody should say anything is the bottom of the well at this time. What he can reasonably say is: when he looks into the well, he cannot see past quantum mechanics. He does not see any reason for there to be anything beyond it. "I cannot see past it" and "there is nothing there" are different sentences. One is a report of vision. The other is a closing of the well.

Do free will, consciousness, and the self connect to a deterministic or probabilistic substrate? Susskind is as puzzled as Fridman. He can, however, define an observer at a technical level. An observer is a system with enough degrees of freedom to record information and become entangled with the thing it is measuring. Entanglement is the key. When an apparatus, or observer — same thing — interacts with the system, it does not just look. It becomes physically entangled, and that entanglement is what we call an observation or a measurement. Does that satisfy him personally? Yes and no. He finds it very satisfying that there is a mathematical representation of observing a system. He is uncomfortable with it in the same way he never gets comfortable with five dimensions. His brain is not wired for it. Four dimensions he can live with a little better, because he can imagine the fourth as time.

Are there echoes of that entanglement at human scale? Absolutely, he says. We are quantum-mechanically entangled with everything in the room; if we were not, we would not be observing it. Space, among the people he is comfortable talking to, they have a pretty good idea is emergent, coming out of entanglement and other things. Time still sits in the equations much as Newton, modified a little by Einstein, would have called it, and mostly it is not treated as emergent. In the microscopic physics, time is completely symmetric forward and back. You do not need an arrow for most of those phenomena.

The arrow appears only when systems are big enough for thermodynamics and entropy to matter. Entropy is not a good concept for a small system. It emerges from large numbers: probabilistic, statistical, thermodynamic, needing lots and lots of little substructures. Do we understand the arrow? He thinks we understand it better than most people think we do. It is a statistical idea. Fling a deck of cards in the air and look: it is random. It does not go from random to simple. It goes from simple to random.

Can the arrow break down? In a laboratory, he says, you can take a system somewhere between small and large and make it go backward. Something that looks as if it only wants to go forward, for statistical-mechanical reasons, for the second law, can be manipulated very carefully to run backward. He does not think you can reverse Humpty Dumpty. In a controlled situation you can stop a system that appears to be evolving toward randomness, reverse it, and send it back. Fridman hears time travel. Susskind says he did not say time travel was possible. He said you can reverse a system's steps, reverse its trajectory.

Small systems have no sense of an arrow; atoms are completely reversible. The second law is a law of large numbers, so you can break it, but it is hard and requires great care. The bigger the system, the harder, because you have to overcome chaos, and that demands more and more precision. Ten particles you might manage with effort; a hundred is really hard; a thousand or a million, forget it — not for a fundamental reason, only because it is technologically too hard. Time travel to the future is easy: close your eyes, sleep, wake up. A good nap gets you there. He would not call reversing a trajectory time travel, because it gets mixed up with what science fiction means.

The picture is an idealized frictionless billiard table. Balls start in a triangle, you break, they fly off. Try to stop every ball dead at some instant and reverse its motion. If you did that with tremendous care, they would reassemble into the triangle. That is a fact. Two balls, probably; three if you are lucky. As the system gets more complicated you must be more precise, because the tiniest error is magnified and you cannot complete the reversal. You are running the program backward, not visiting last Tuesday.

If the world is a program, how hard is a computer that runs the universe? Susskind's answer splits in two. There are mathematical universes we know, one of them anti-de Sitter space. A classical computer can only solve its equations, not make it work like the real system. A quantum computer, big enough and robust enough, could probably simulate a small version of an anti-de Sitter universe and treat the center as a kind of cosmology. The trouble is the universe we live in. Ours is de Sitter geometry, and we do not really understand its quantum mechanics at all. At present he would say we would not have the vaguest idea how to simulate a universe similar to our own.

They differ by the sign of a single constant, the cosmological constant: one negatively curved, one positively. Anti-de Sitter, the negatively curved one, you can think of as an isolated quantum system in a box with reflecting walls. De Sitter is the one we actually inhabit — exponential expansion, dark energy, whatever we want to call it — and we do not understand that mathematically. Not everybody would agree with him, he says, but they would definitely agree that *he* does not understand it. On origins, his favorite theory is eternal inflation. Understood correctly, he thinks it would be infinite in both directions. Fridman assumes he likes it because it is mind-blowing. Susskind's reason is plainer: we do not need a beginning. How do you think about infinity? He has just had this conversation with Sergey Brin. "Sergey Brin is infinitely rich. How do you test that hypothesis?" There is no operational test. Whether infinity has a place in physics he will only call a very good question.

The Event Horizon Telescope's first black-hole image he calls an incredible triumph of science. That black holes collide is not a surprise; they seem to work exactly as they are supposed to. Will we learn a great deal? Maybe, but not really about black holes. Why nature has holes at that particular mass scale, and why they are so common, may tell us about the evolution of structure in the universe. Go back a hundred years: general relativity, the discovery of black holes, LIGO, technology he would not have believed 30 or 40 years ago, measurements on a scale of 10 to the minus 21. The picture did not change how he visualizes holes. It confirmed. Direct observation that Einstein's gravity works at the level of black-hole collisions is awesome. He knows some of the people involved. They are ordinary people. That they could carry it out shocks him. Little *Homo sapiens*. Fridman offers slightly advanced lemurs. Susskind lets it stand.

Which questions can science not answer now but might soon? Consciousness, he has already said, is somewhat within reach, and in the hands of computer scientists and neuroscientists, not physicists. Physicists will try to simplify it down to methods they know, and those methods may not be appropriate. What he finds awesome is another path: not only machines that learn, but machines that evolve their own architecture, only partly under our control, under the control of machine learning. Google taught computers to play chess not by teaching them chess but by having them play against each other. That is a form of evolution. The machines evolved intelligence without anybody telling them how, not engineered, just playing and getting better. That makes him think machines can evolve intelligence. What exact kind, he does not know. Getting better and better may give better clues to what goes on in our own.

Last question: what can science not answer, and maybe never? Is there intelligence underneath the whole thing — you can use the G-word if you want. Are we a computer simulation with a purpose? Is there an agent responsible for the whole thing? Does that agent satisfy the laws of physics, the laws of quantum mechanics? Is it made of atoms and molecules? The questions seem real to him. Some philosophers would say a question is not a question unless it is answerable, and this one does not seem answerable by any known method. He still thinks it is real. That is where Fridman stops. Across the hour Susskind keeps doing the same move: give the technical definition, then admit the definition does not sit comfortably in a three-dimensional brain. Observation is entanglement, and he is still not wired for it. Time is statistical, and he thinks we understand it better than people admit, and he will not pretend that settles the well.