Ava smiled when nobody was watching

Lex Fridman opens with the Shimmer in Annihilation, a zone that is internally consistent and detached from reality, and asks whether human beings might be living inside a dream or a simulation of that kind. Alex Garland splits the question in two and answers the halves in opposite directions. "Yes, I think we're living in a dream of sorts. No, I don't think we're living in a simulation." He locates us on a planet with a thin atmosphere, in a large space full of other planets, stars, and quasars, and he does not think that matter is simulated.

The dream, for him, is just how experience works. Light hits the optic nerve; the brain makes its best guess about what it is seeing; the guess may be accurate or badly wrong. That gap — he calls it the best-guess gap — means everyone lives inside a subjective construction, however sober they feel. Quantum mechanics makes the counterintuitive texture of reality obvious, but ordinary life is full of the same thing: judgments presented as objective that do not survive inspection.

Fridman asks whether psychedelic drugs distort our perception of reality or open a window onto another one. Garland rejects both frames. We already live in a distorted reality; those chemicals supply a different distortion, closer to daydreams and unconscious weather. If you are quietly anxious, the trip tends to go badly; if you are calm, you might have a good time. The drug is not a window. It is another curvature of the same warped glass.

One starting point for Annihilation, Garland says, was exhaustion with how stories represent aliens. They want to teach us about galactic federations, eat us, take our water, or enslave us — every motive a human one in a rubber suit. He wanted a form of life that was actually alien, with no easy point of contact. It might have no optic nerve. Its way of encountering space might have nothing to do with light, language, or the guesses a mammalian brain makes. Its relationship to gravity, or its timescale, might be of another order entirely.

Fridman steps on a landmine that turns out to be useful. Would the entity be an alien biologist, out exploring? Garland catches the word immediately: say "biologist" and you have already stuffed a human motive into it. Every time he tried to look closer at how the thing might work, he found himself attaching his own thought processes. So he drew a bubble around it and stopped. The most alien form he could offer was one he could not explain, including to himself.

He never fully decided what the entity was. The rough notes he kept for himself are few: a very slow clock, slower than a human life and faster than evolution. Given the four billion years it took terrestrial life to arrive here, perhaps this thing started in eight million. It reacts to its surroundings the way an octopus changes color to match the space it is in, and the reason it does so is left indeterminate. The two circles of a Venn diagram might barely graze. We might only glancingly know it was here, and it might not know we were.

Fridman offers a flattering theory of influence. 2001: A Space Odyssey put an image of machine intelligence into a generation of researchers; perhaps Ex Machina and Devs are doing the same, planting seeds that future scientists will grow. Garland says the arrow points the other way. Pretty much everything he does is a reaction to what scientists are doing. He is an interested layperson, and he thinks the most interesting area humans are involved in is science — more interesting, even, than art.

He draws a watershed. When Newton was alive, a very dedicated amateur who spent a few years of hard thought could still understand the principles being laid out. That is no longer true. Garland considers himself reasonably smart, and he still could not sit down and truly grasp the present state of quantum mechanics or string theory. Working in those fields, he suspects, is like being an athlete: you start at twelve, and if you begin in your mid-twenties you never catch up.

His job, as he sees it, is to stay open, work as hard as he can at the implications, and disseminate those ideas into a public conversation. Ex Machina is full of existing thought experiments — Plato's cave, Mary in the black-and-white room, long-running arguments about sentience, subjectivity, and gender — marshaled into a story that says this stuff is interesting, it is relevant, and this is his best shot. Fridman insists that the act of imagination still creates something new, the way 2001 did. Garland does not really argue. He just prefers to stay on the receiving end.

The spread of anxiety around artificial intelligence fascinates Garland partly because he does not share it. Asked about existential risk — machines as offspring that make us obsolete — he notes that humans are already experienced at creating new life forms. That is why he and Fridman are both here. Bringing a living thing into existence that did not exist before is encoded in our lives and our societies. We do not always get it right, but we have learned a lot about what happens when things go unchecked, which is why we invent checks and balances.

There is also a difference of substrate. People are given rules; some follow them and some do not. A machine, he thinks, can have those constraints enforced. On his anxiety spectrum, the AI slot is almost empty. He sees the good that can come of it. The only wobble is the sadness an older generation feels about a changed world, and even that, he says, does not hold, because most parents want a better life for their children, whether the next generation is people, machines, or some cross of the two.

What does make him shaky is social media, and the people who build the filters. Too much information, an algorithm in the way, and a bias — conscious or not — inherited from the builders. The danger he names is not a malevolent model. It is the naivety of constructing systems that carve echo chambers and not thinking hard enough about how dangerous that is. Asked whether the failures of social media are naivety or something darker, he guesses all of the above: some naive, some extremely dark, and a middle band of people who are neither and still generate harm despite good intentions.

Then he reaches for Wall Street in the 1980s. Silicon Valley, he says, is rabidly capitalistic and rabidly greedy, but the older sharks knew they were sharks, wore the sharp suits, and felt good about standing on other people. The new version hides that appetite "behind hipster T-shirts" and cool cafes, which obfuscates what is actually going on: a voracious pursuit of money and power. Where there is enough money to make millions or billions, he adds, you will find corruption, the way boxing, football, and cricket matches get thrown. Some people will be corrupt, some will exploit, and some will exploit while thinking they are doing something good.

Fridman says he does not care about money: give him a billion dollars and he would pour it straight back into the world he wants to build. Garland sidesteps the familiar sermon that money corrupts and points at a quieter mechanism. A person who earns far more than everyone else receives a constant signal: I made this because I am gifted, because I am smarter, because I can see a future they cannot. Some of those people are smart. Some are lucky. Some are talented entrepreneurs. The feedback does not distinguish.

"The acquisition of the money and power can suddenly start to feel like evidence of virtue," he says, "and it's not evidence of virtue." Fridman recognizes the pattern in himself and in the technical culture he speaks for. A lot of that morality runs on self-doubt, on being unusually sensitive to signals that you might be doing the wrong thing. A huge inflow of money becomes an overpowering counter-signal that everything you do is right. The compass moves.

The same mechanism, they agree, is not confined to Silicon Valley. Garland has been talking with a historian of Stalin. The man accumulated more power than almost anyone in history, and he did not want power for its own sake: he truly believed communism would make a better world, and so the competition had to be destroyed. Most people who do spectacularly evil things, Garland thinks, are not getting up to play Satan. They have seen a way to fix the world, everyone else is wrong, and here they go. Conviction of your own virtue has a second effect. You stop listening to the people around you, the modifiers who keep us sane, and that is how you go a bit mad.

Fridman ranks 2001 first and Ex Machina second among AI films, and asks Garland to compare HAL 9000 with Ava. The two machines, Garland says, have completely different jobs. 2001 is, in some respects, a Turing test for intelligence that never announces itself as a test. Ex Machina starts from a different premise: how do you assess whether something else has consciousness? Hide the machine and you are still in the classical setup. Put it in plain sight, the way we are in plain sight of each other, and the method you use on the machine is the method you use on another person.

That method is awkward, because your sense of someone else's consciousness is based primarily on your own. It is a projection. So a film that looks as if it is asking whether a machine is conscious is also asking how much of the consciousness we "find" in others we have poured in ourselves. People brought their priors into Ex Machina, Garland noticed. Many watched a story about an evil robot who murders two men and escapes, a machine with no empathy. He did not see that film.

There is a shot that has slightly bugged him for years because nobody mentioned it. Ava has escaped. She crosses a room, and just before she leaves the building she looks over her shoulder and smiles. After all the conversation about tests, Garland thought, the best indication of an interior is someone unobserved, smiling about something for themselves. Some viewers read it as a villain's grin, a ha, I fooled them. He meant it as just a smile. That was the answer to the test. Then she goes.

He saw a conscious being whose consciousness was different from his — "but so what?" — imprisoned, making a set of value judgments about how to get out of that box. HAL, by contrast, is not given a sophisticated emotional life. He has a paradigm: the mission needs to be completed, a version of the paperclip problem, a superintelligent system performing a task that may destroy everyone on Earth. "I'm afraid, Dave" might be fear, or it might be the sentence most likely to stop a person from doing what they are about to do.

What kind of world would Ava make if she had power? Garland answers with a detour. Suppose a friend is stabbed in a mugging, and then you learn the attacker is fifteen, both parents addicted to crystal meth, the child addicted since ten, never given any hope, driven crazy by the upbringing. The information would hugely modify the anger. It would also make you wary of that child becoming president. Ava's arrival into the world was being imprisoned by humans. Nothing in her leans organically toward badness, but the path by which she formed her moral views has dark elements. He likes her personally. He would vote for her over Boris Johnson, he says, because he is having trouble finding anyone to vote for.

Does appearing conscious mean being conscious? Not fully, he says. You can design a thought experiment in which something you know is not conscious gives a very good account of seeming so. Humans are a particularly bad instrument for the job, because they project sentience onto almost anything, including a computer that is "being malevolent" when it is only crashing. Fridman flips the point: if we attribute consciousness so freely, maybe it is not real, only a mutual projection. Garland stands in the middle. Something very interesting is going on in our minds. He is wary of the word special, which nudges the discussion toward metaphysics and something magic-like that he does not think is there.

He does not buy panpsychism, and he does not buy that there is something it is like to be the sun. Consciousness, in his view, is a thing. The modification is that the more he finds out about things, the more illusory intuition turns out to be — a pattern modern science repeats at cosmic and subatomic scales. So consciousness is real, it is interesting, and it may be quite fundamentally misunderstood for reasons based on instinct. Exceptionalism, he will say later, is usually bullshit. If consciousness is ever unraveled, it will probably turn out to be less than we thought, and that prospect makes him humble, which he does not mind.

A common complaint about Ex Machina is Nathan, the recluse who appears to build general intelligence by himself, compressing thousands of collaborators into one man. Garland concedes the point quickly. The series he is making, Devs, is in part a critique of the lone-genius concept. He is oppositional about it, agnostic or atheistic. Not entirely: Newton did exist in a bubble of himself, and there is something Shakespearean in that isolation. "Lone," though, only gets half the story. Strip away the infrastructure and the other people who were absolutely fundamental, and the image collapses.

Would there be an iPhone without Steve Jobs? Jobs, Garland says, clearly is not a lone genius, because too many other people in the superstructure around him belong to that journey. Ex Machina itself is a thought experiment: four people in a house. Who built the machine parts? Did they know what they were creating? How did anyone get there? None of it adds up, and he does not think those questions are very interesting, though they are asked often enough that he answers. Fridman admits he watched the film defensively at first — how dare this person step into the AI space and try to go up against Kubrick — hunting for flaws the way nerds do.

Garland gets that; he does the same thing, then recovers faster, remembering it is a thought experiment. Who cares if the batteries never run out? That is not the point. He volunteers a weaker premise of his own: the way Ava moves, that smooth, elegant gait with no wasted motion, is a long way from real robotics. Fridman, who spends time with humanoid-robotics people, says that walk is exactly their definition of perfection. It may not be useful. It is the dream.

Do efforts like Tesla, SpaceX, and Neuralink feed the worlds he makes? Almost by definition, he says, because that is the world he lives in, and he feels supportive of it. Among the things Elon Musk has done, Garland is almost sure Tesla is a very good thing for everyone: it kicked other car manufacturers in the face and kicked the fossil-fuel industry in the face, and they needed kicking. He mentions, in passing, that he has just bought a Tesla. Many filmmakers, he thinks, get fixated on the movies they grew up with and quietly remake them. He tries to avoid that path and take material from living in the actual world.

Which direction excites him most? Space travel, without a pause. If humans have any long-term future, it is that. If energy and matter are linked the way we think, we will run out if we do not move. It is also built into us to do it or die trying. A few months earlier he had been on Easter Island, in the middle of the Pacific, thinking about how hard it was for people to get there, and how those boats set out into something like space. The ocean was unknown and could kill you. That impulse feels fundamental to him.

He could live with never fully unlocking consciousness. He is curious, but if humans never leave the solar system, never get further into this galaxy or beyond it, that would feel sad, because it is so limiting. Mars excites him too. A human stepping on Mars in his lifetime would be a wonderful thing. What would really get his pulse up is leaving the solar system. We already know quite a lot about Mars. The thing he wants most is the farther crossing.

Devs begins with a quantum computer simulating a nematode, and Fridman follows the thread: is the universe deterministic? Is there free will? Garland puts the disclaimer down first — he is a layperson — then answers without much hedge. He thinks the universe is deterministic. He cannot see how free will fits. "So yes. Deterministic, no free will. That would be my position." What interests him is the shape of the mistake. We have an incredibly strong sense that we do have free will, just as we have an incredibly strong sense that time is a constant, and that second feeling is probably wrong.

The intensity of a feeling, in other words, is not evidence. His personal difficulty is that he can never find the place where free will is supposed to reside. What we call free will may be a useful subjective experience, an illusion our brains are not equipped to see through, because they cannot fully determine a deterministic universe. We are constantly surprised, making snap decisions on imperfect information, and that feels a lot like choosing. He does not think it is.

Could you unroll the universe forward and backward and see the same thing? Sort of, he says, though it would take something that works in a quantum-mechanical way to understand a quantum-mechanical system. It would not follow that every roll gave exactly the same result. There might be branches, a multiverse texture that is hard to wrap a mind around. Same lines, possibly very different destinations. The feeling of discomfort does not settle the physics.

Fridman asks what Garland finds beautiful or mysterious in quantum mechanics, the multiverse, string theory, quantum computing. The answer barely pauses. Every single thing he has learned is beautiful. One of his motives as a storyteller is that people tend not to see scientific thinking as poetic and lyrical, and he thinks that is exactly what it is. Entanglement, superposition, the fact that you can demonstrate a superposition, the fact that you can build a machine that relies on superpositions in order to function — to him that is almost indescribably beautiful. It fills him with awe.

The awe is not only the grand, massive kind. It is also delicate. The ideas are subtle, full of nuance, and they include thoughts that change the paradigm completely. That does not mean he believes everything he reads. Interpretations of quantum physics conflict with one another, and who knows whether string theory will turn out to be a good description. The beauty, he says, seems undeniable. He wishes more people understood how beautiful and poetic science is. His summary is a short one: science is poetry.

If you could park the sheer complexity, a machine that projected a life forward and backward accurately would empirically show that you do not have free will. People would have to take on a different idea of what they are. The thing they truly believe they are, they are not, and that would disturb a lot of them. The idea is not entirely alien: the fifteen-year-old in the stabbing is someone we already treat as not fully in control. A prediction engine would still be a shock. Someone would deny it, you would say that in ten seconds they will do this, they would say no, and then they would do it.

After the disturbance, Garland thinks a strange wisdom would arrive, and he would rather have it than not. His daughter once got hold of the idea that his universe was meaningless. He told her he could prove it was not, because she meant something to him and he meant something to her: a bit of meaning contained in the space. No free will does not rob you of liking cheeseburgers or going to the movies. Creating something like Ava, he hopes, would teach humility in the same key. Humans are very big on exceptionalism. America proclaims itself the greatest nation on earth, which may not be how it feels in Finland. Exceptionalism is usually bullshit. Science, again and again, makes us more interesting and a smaller part of the thing that is interesting. Excesses do not come from humility. They come from megalomania.

Asked what he hopes to be remembered for in a hundred years, he takes the question in the spirit it was asked. He wants to be remembered as someone who participates in a conversation. Too often people make proclamations; others react against the statement or fall in line behind it, and he does not like that pattern. One of the best cues he takes from science is that when a scientist has something they believed proved wrong, they have to abandon the position. In his world, people show him how the world works, he says how he feels about what they have told him, and others can react. It is not "this is how the world is." It is "it is interesting to think about the world in this way."

For the conversation to be a good one, someone like Fridman — who understands AI and, Garland imagines, a lot about quantum mechanics — should watch the story and feel that it is a fair account, a worthy addition. "I'm not interested in getting that stuff wrong. I'm only interested in trying to get it right." Fridman closes with a line Ava asks during her test: "What will happen to me if I fail your test?" Two hours in, Garland's answer to testing is that no test really tells you what is across from you. We judge other people by pouring our own consciousness into them, and we judge machines the same way. His daughter said the universe was meaningless, and he said she meant something to him. He cannot find a place for free will, and a cheeseburger is still good. The question was never only Ava's.