A cheap simulation would render only you, and the wall would have nothing behind it

Fridman starts with a question he admits is unanswerable: what matters more, understanding the universe at a fundamental level or understanding the human mind? The glib reply is that the brain is part of the universe. Carroll will not take it. Once you know particles and forces, do you understand the mind? "No, certainly not. We cannot understand how ice cream works just from understanding how particles work." He is a believer in emergence. Tables, chairs, planets, and people are not spoken in the language of particle physics.

He does leave the question a little dignity. There may be general principles of complexity, biology, information processing, memory, knowledge, and creativity that go beyond this particular brain, and those could count as understanding the universe. The human brain is still, as far as we know, the most complex thing in it. Expecting the fundamental laws to yield direct insight into how it works is, in his word, absurd.

Does he think of the universe as a computational device? The honest answer is no. Information processing clearly happens in it, and in some sense it is like a computer, but Carroll has argued on his blog that the universe is more like a computation than a computer. It happens once. A computer, even a pocket calculator, is a general-purpose machine set up to answer a class of questions. The universe is not that. Processing is something that goes on inside it, not what it is.

Seth Lloyd, an MIT colleague, feels otherwise. Carroll recounts a panel: Lloyd gave his universe-as-computer pitch, and Martin Rees asked what is not a computer. Lloyd said it was a good question and he was not sure. If the definition is wide enough, everything qualifies and the analogy loses force. Is the Moon going around the Earth performing a computation? You can write a definition that says yes. It is not a useful one. A comparison earns its keep by excluding something.

He still finds the information-processing picture useful in the right context. He even wrote a paper called "quantum circuit cosmology," modeling the whole universe as a quantum circuit whose qubits become more and more entangled. That model is his entrance to a puzzle he thinks is so deep almost nobody talks about it. Space expands. In a region of space the number of degrees of freedom — ways quantum fields and particles can arrange themselves — is arguably finite. If space gets bigger, do you get more degrees of freedom? Infinity times two is still infinity. A finite number means the universe is not closed: new degrees of freedom have to come from somewhere.

The suggestion in that paper is that the extra degrees of freedom were always there, just not entangled. The three-dimensional space we know is the entangled ones making up spacetime. As the universe expands, a pile of qubits sitting in their zero state get entangled with the rest through those circuits. Spacetime, on this philosophy, arises from the entanglement of fundamental quantum degrees of freedom. At the Big Bang almost all of them existed, unentangled, which is what a low-entropy, very small, very simple beginning looks like. Expansion is more and more of them joining the world's tangle.

Fridman asks about Nick Bostrom's thought experiment that we live in a simulation, and wants it treated as an AI question: how hard would our world be to fake? Carroll grants physical possibility and denies that we actually are in one. He wants to be a good Bayesian: under the hypothesis, what should you expect to see? First, not a universe this big. The observable universe holds about two trillion galaxies with some 200 billion stars each. Lighting all of that up to run a simulation looks like a waste of resources.

Second, not this resolution. Finite-resource simulators would not render the whole sky. The obvious next step is that only you are being simulated fully. Everyone else is a non-player character, a chatbot. You see the wall; there is literally nothing on the other side. That is what an efficient simulation of a person would look like, and it is not what we see. Saying we would not know what a simulation looks like, he argues, is abandoning your Bayesian job. Your job is to say, under this theory, here is what you would expect.

Bostrom's population argument — if simulated beings are conscious, most conscious beings will be simulated, so you probably are one — Carroll thinks must be wrong. If we can simulate others, they can simulate others, and the chain hits a bottom because our physics has a bottom; we are made of atoms. There will be a cheapest possible simulation, and the original logic says you should live there, because that is where most people are. We do not look like 16-bit creatures at the edge of resolution. Much simpler things would be easier to make. He also distrusts the anthropic habit of treating ourselves as typical observers in the whole universe. You can judge typicality inside what you already know. You cannot place yourself in the whole.

Is there intelligent life out there? Carroll's guess is that there is none in the observable universe besides us. The likely numbers, he says, are zero or billions. If there were billions, we would have noticed. A dozen Star Trek civilizations in the galaxy, not a billion, strikes him as bizarre. Zero is easy to imagine: a bottleneck from single-celled to multicellular life, or from life to technology, that almost nobody gets through. Billions that have remained hidden is hard.

Fridman asks whether intelligence could live on other size and time scales. Carroll wants humility: we do not even agree on what life is, let alone intelligent life. Dolphins may be intelligent and never invent space travel, because they live in the ocean and have no thumbs, so no smelting, no technology. The universe could be full of clever species that never make gadgets, and that is compatible with the data. Interstellar molecular clouds, neutron-star surfaces, galaxy-scale things whose heartbeat is a hundred million years: he is open-minded, and also notes that we all share the same physics. Atoms have a size, stars have a lifetime. There might be a sweet spot in meters and years that we happen to sit in. He will not mind either way.

The search, he says, has been done in the dumbest way available: pointing radio telescopes at the sky. Why would a super-advanced civilization wastefully beam radio in all directions for millions of years? If you thought a solar system might someday grow minds, you would not hail it with radio. You would send a spacecraft and park it, a monolith waiting, 2001-style. The converse holds: there may already be artifacts in our solar system, and we have not explored thoroughly enough to find them. We do not think that way because we are young and impatient. A probe to another star and back is longer than a lifetime. On hundred-thousand-year or million-year scales, it is "clearly the right thing to do."

SpaceX and space travel still matter to him, including other star systems. People overestimate the difficulty because they assume today's lifespan. Science fiction invents faster-than-light so the plot will not bore the reader. We will not beat light, but we can imagine lives stretched to thousands of years, at which point the stars are effectively closer and a hundred-year trip is nothing. A friend who thinks about this told him: you and I will die, but maybe not our grandchildren. Predicting is hard; spreading humanity and biology off a threatened Earth is still an important long-term goal, and a race against time.

The scientific mystery he thinks is within reach is the origin of life. Life-as-we-know-it, which we can discuss with some intelligence, needs three things: compartmentalization, a membrane around the cell; metabolism, taking in food and turning it into work; and replication, passing information about what you are to the next generation. In the lab, lipid bilayers that form little walls are pretty easy. Replication is close: the state of the art cannot yet make one molecule that copies itself, but it can make two that copy each other. Metabolism is the hard one, even though it sounds obvious. The cellular machinery in our bodies is complicated enough that it probably did not pop into existence, and it probably took a long time and many wrong turns.

If he ran the NSF he would flood the area with money. Making life in the lab, and understanding how it started on Earth, would change our view of the world and throw off practical effects, including on disease. Synthetic biology is a huge frontier. Craig Venter's artificial cell — DNA pulled out, entirely new DNA put in, the cell booted — was a tremendous accomplishment and still only a swap of an existing startup program. Getting a metabolism running from nothing is the remaining wall. The leap from that to artificial consciousness, Carroll does not think we are on the threshold of. We understand so little of what consciousness is that it is hard to believe we are about to build it. No obstacle in principle; not soon.

Fridman suspects that once we have intelligence we will be surprised how easy consciousness was, because it is something we all fake. Carroll likes the idea: we are, at every moment, less conscious than we think, and putting an AI in a robot body would help the faking a lot. A Roomba that says it is conscious and will not be talked out of it is a joke that is also what humans do with each other. Consciousness and intelligence are social constructs as much as they are chains from electrons to neurons. Kate Darling's simple animal-like robots look like they are in pain, and people react badly. You want to say they are just electrons going around. Then you remember that you and I are just electrons going around, and that is what pain is. He can easily imagine a spectrum from those toys to a human: things that fail a strict Turing-test definition and still walk and talk as if they were conscious. Siri is already close. We will not one day shout aha. We will creep up on it.

Photos you already cannot trust; video is catching up. The last defense is a physically present person, and faking that may not stay hard. Carroll finds the conversation depressing. Fridman finds it exciting. Carroll's line is that we are in a phase transition and people in the middle of one are terrible at imagining the next fifty years. Fridman is not blind to threats; he is excited by technology's power to meet them, and he has found brilliant people to be good people, so the army of the good looks large. Carroll is far less optimistic. Goodness and badness, he thinks, are equally distributed among the intelligent and the unintelligent. He sees little correlation. Neither has proof. Opinions, he says, are free.

What can science never answer? The obvious one, Carroll says, is what is good and bad, right and wrong. Science tells you what the world is and what it does. It does not say what the world should do, or what we should do. We are part of the world and we can feel that something is right or wrong. Moral philosophy, to make a long story short, is systematizing those intuitions. Science might predict what we will choose. It will never judge whether we should have.

Fridman had thought science was supposed to reach outside the silo, then found academia extraordinarily partitioned, and asks how Carroll prepares for interdisciplinary conversation and how he would advise others. Carroll tells a trainer story. Someone asked a personal trainer how to psych yourself up to work out. Why are you asking me, the trainer said; I cannot stop. Carroll's reaction to "how do you do interdisciplinary talk" is the same. It is what makes him go. He did it long before anything was recorded. One reason he started a podcast was to force himself through the books he had bought: if Patricia Churchland is on next week, her book gets read this week.

Then the less pretty part. Academia is extraordinarily siloed. People rarely talk across fields, and when they do it is punished. The ones who succeed at it generally first became very successful inside a little silo, and only then expanded. He is candid with graduate students. Most of them will not become faculty. Live the life you want, eyes open about the job-market cost: the broader you are, the less time you spend hyperspecializing, the lower your chances. He thinks that is terrible. He does not like it. It is the reality. Plenty of wonderful scientists have zero interest in branching out, and that is a fine other life. What disillusioned him was the romantic picture of intellectual academic life. The idea that reaching beyond your discipline is a positive good is so rare in universities that it may as well not exist.

Fridman offers the workaround he is living: without Carroll's podcast he might not be having these conversations at all. Papers are exchange of ideas, but interdisciplinary papers are much harder. Conversation is easier, so conversation is a beginning. AI, he says, should be looking at physics, psychology, neuroscience, and broader impact, not only computer-vision contests and a few datasets. Carroll thanks him. Fridman adds that you never know how the world changes: graduate students come up inspired, some will become faculty, ripples go out. They end on that balance of pessimism about the institution and optimism about the people it failed to contain.