Maybe it’s best to start at the end.
“I’m sorry, I’m running a bit short of time,” I say casually on the phone to Robin Hanson before the last question of my last interviewee. “Is there anything we haven’t covered that you think is important to say about this?”
Hanson, who is, among other things, a professor of economics and a research associate at Oxford University’s Future of Humanity Institute, speaks extraordinarily fast. Throughout our interview, his answers have tumbled out before I’d even finished asking the questions. Transcribing our conversation, I had to drastically slow down the recording to keep up with him. But at this question, he took a long pause.
“Uhh,” he says, drawing the word out into almost a sigh. “This topic — this topic breaks people.”
Have you started to get the feeling lately that nothing is real? That the world seems to be testing your credulity or cracking around the edges? You’re not alone.
Recently I’ve had this weird feeling, kind of like I’m the eponymous character in The Truman Show, in the sense that I feel like I’m living in a world that seems real, but one that I can also sense cracking around the edges, failing to connect at times in unsettling ways.
We’re not the only ones questioning the nature of reality.
Journalists covering the phenomenon of Trump and the ecosystem of the internet’s darker corners, myself included, have retreated into a sort of gallows humor. “LOL, nothing matters” became a common nihilistic refrain as, time after time, the reality we reported on seemed to come unmoored from all the rules of cause and effect. It started to feel not just like we were living in a fictional world, like a video game, but also that it wasn’t even particularly well-written.
As late-night host Stephen Colbert said on his show in July, “As I was looking at the news today, I went: What is this feeling I have? This feeling I have is that this is so weird that everything seems weird… that weirdness leeches into all the news, that feeling of weirdness, and after a while I’m not sure what’s real or not.”
We’re not the only ones questioning the nature of reality. A strange and fascinating argument is playing out, one that has set quantum physicists against philosophers and eccentric tech billionaire Elon Musk. They are arguing about the surprisingly plausible idea that our reality isn’t really real at all, but a computer simulation.
The idea rocketed into the mainstream public consciousness in 2016, when Musk endorsed it on stage at a tech conference.
According to a line in a New Yorker article profiling Silicon Valley incubator Y Combinator, “Many people in Silicon Valley have become obsessed with the simulation hypothesis, the argument that what we experience as reality is in fact fabricated in a computer.” Tantalizingly, the article adds that “two tech billionaires have gone so far as to secretly engage scientists to work on breaking us out of the simulation.”
The key paper for simulation theory, the idea to which Musk was referring, is a now extremely famous 2003 paper by an Oxford philosophy professor named Nick Bostrom, the founding chair of Oxford University’s Future of Humanity Institute.
Bostrom’s core idea was a compelling one. It has been amended, and different forms of it have sprouted, but in its most basic form it goes like this: First, assume humanity makes it in the future. If we do, computing power will likely increase; the more powerful computers get, the closer these theoretical future beings will get to the point where they are able to simulate worlds with conscious beings like us in them. Then, evaluate how likely it is that, across the whole span of time, we are the ones living in the real original world rather than in one of those simulations of it?
At the beginning of our interview, I ask Hanson, Bostrom’s colleague at the Future of Humanity Institute and author of How to Live in a Simulation, if this is an idea I’m supposed to take literally, or if it is more of a thought experiment.
“Certainly Nick Bostrom’s version, the version of people I know, are taking it seriously,” he says. “It’s a real possibility. It’s a matter of what probability I assign to it — but it’s certainly way above zero. Even well above one in a million.”
“As I understand it,” I say, not understanding it, “is that it’s overwhelmingly the likely probability. That’s what he’s saying, right?”
“I’m sure if you asked him straight out, he’d probably give you a probability,” he says. “I would give you a probability.”
“And what would you—”
“I’d give it over one in a million,” Hanson says carefully, “and less than 50-50.”
But Bostrom’s theory, as espoused by Musk et al., has come under attack in recent months. In a paper published in the journal Science Advances, two theoretical physicists also based at Oxford address the question of whether reality is a simulation more or less directly, and they more or less definitively proved that it was definitely not.
The two physicists, Zohar Ringel and Dmitry Kovrizhin, found that on a purely technical basis, the computing power required for creating a simulation down to the quantum level isn’t merely huge, it’s preposterously vast. Impossible.
At least, that’s how it played out in the media. “Physicists Find We’re Not Living in a Computer Simulation” was the headline in Cosmos, adding, “The sci-fi trope might now be put to rest after scientists find the suggestion that reality is computer generated is in principle impossible.” Futurism.com went with “Sorry, Elon. Physicists Say We Definitely Aren’t Living in a Computer Simulation.”
“My problem is [it’s] not a theory, it’s some conjecture,” Kovrizhin, one of the theoretical physicists who authored the paper, told me about Bostrom’s simulation theory. “It’s based on an assumption, and the strongest assumption is he thinks if an alien race is powerful enough, it will have nothing to do but generate infinite simulations, and I don’t know what that assumption is based on.”
“It’s very hard to prove or disprove him. But what I’m trying to do is try to do an honest job in theoretical physics, when we apply physical law to our reality,” Kovrizhin said. “We can’t say that we proved we are not living in the matrix. What we can say is that with certain models on a classical computer, it would be hard to simulate certain systems; the number of particles in the universe is very large.”
In terms of scientific discourse, that’s some pretty vigorous pushback.
“My original understanding of the theory, which I think is flawed in the same way that the particle physicists’ understanding is flawed, is the idea that any simulation would have to be on the scale of the universe,” I suggest to Hanson. “And I think the paper that came out was saying the universe is not a simulation because—”
“Well, that’s one mistake to make,” Hanson cuts in, “making the assumption that any simulation would be on the scale of the entire universe all through history.”
“But their mistake seems to be the sheer impossibility of enough computing power simulating every particle in the entire universe,” I say. “Simulations would be humans interested in humans, not a universe A.I. interested in the universe on a universal scale, right?”
“Right,” Hanson says. “The simulation of history is intended to be focused on pivotal actors and pivotal choices.”
Jim Gates asks me to call him Jim. “I am a simple country theoretical physicist and am not at all enamored of titles,” he says in his reply to my email requesting an interview.
“I want to be very clear,” he wrote. “I am not a supporter of the idea that our reality could be a simulation.”
Gates is a professor of theoretical physics at Brown University who served on President Barack Obama’s scientific advisory council. “The main problem with the so-called simulation hypothesis is that it fails to meet a requirement of any proposal claiming to be a scientific one,” Gates said. The stern exasperation of his tone was only slightly belied by the fact that the entire message was written in Comic Sans. “There is no way to falsify this idea.”
“It may surprise you to hear this is my view,” he went on, “as there is a vast amount of online content dedicated to the idea that my own research has driven me to be a supporter of this proposal.” Gates’ impatience at the question was, in hindsight, understandable. His landmark work in theoretical physics has been, it turns out, somewhat co-opted by the simulation crowd.
Such a simulation would need to reproduce not an entire universe, but just enough of it to trick its inhabitants into thinking their universe was infinite.
Gates had discovered “error-correcting codes” in the fundamental equations of string theory, and many had pointed to that discovery as evidence of the programming activity of the type of programmer Bostrom’s theory described. On the internet, fans started connecting dots; code became “computer code.” But Gates himself believes that the idea that we might all be “sims” is “fun speculation” but unscientific.
In fact, the problem for both Gates and Kovrizhin — the problem that has led to all this interdisciplinary sniping — is a fundamental one: The theory is not falsifiable, which is to say, it cannot by definition be either proved or disproved. “This is a question that I thought was settled by Descartes back in the 1600s,” Gates said. “His famous conclusion was ‘I think, therefore I am’… not ‘I think therefore I am simulation.’”
But I found myself becoming dissatisfied with the physicists’ approach to the question. Kovrizhin had shown that it was impossible for a universe to contain a computer powerful enough to contain a total replication of itself. But it seemed to me that my universe was much smaller than that. What did it matter if, on a fundamental level, reality itself was untrickable, I found myself thinking, if that whole question can be circumvented by tricking only me?
“It strikes me,” I say to Hanson, “that a future intelligent descendant could be just as interested in abstract experimenting through simulation, how a consciousness reacts to various stimuli, rather than trying to enact history on a wider scale.”
“The simulation argument is based primarily on imagining historical simulations,” Hanson says. “But you could, of course, imagine lots of other types of simulation and imagine you’re in one of those.”
“And, just a back-of-the-matchstick calculation,” I say, meaning a back-of-the-napkin calculation, “what do you think? Does that change your mathematics at all if you include those kinds of simulations?”
“Again, for the generic simulation, the simplest calculation is how many real creatures are there and how many creatures in simulations are there,” Hanson says. “That ratio would be huge, right? There would be far more of the first sort than the second.”
This sounds implausible to me. “There are video games right now that have A.I. characters in them,” I say. At what point are we measuring the consciousness of the creatures in these simulations?”
“You are assuming that these characters in the simulation are conscious enough that you could be one of them,” he says.
“Right,” I say.
“So, if there is a simulation out there with an inhabitant that is not conscious — well, then, that couldn’t be you, because you’re conscious.”
“Right,” I say again.
“So, from the point of view of guessing what kind of simulation you could be in, you are, of course, limiting yourself to simulations where the things in them are conscious,” Hanson says.
“Is there a baseline test for consciousness that I could run on myself?” I ask. “Or is it just cogito ergo sum?”
“Uh,” says Hanson, momentarily nonplussed. “The literature overwhelmingly allows you the strong presumption that you are.”
“Ha,” I say. “Right, because—”
“And, in fact,” Hanson continues, “people have made that the axiom of the hill they will die on: ‘I know I am conscious, and I’m sure about that, and that’s the end of the discussion.’”
“I find myself only… only pretty sure that I am conscious,” I say slowly.
One problem with the theoretical physicists’ debunking of simulation theory is that it is incomplete. It rejects the possibility that our universe is a grand-scale simulation of the entirety of itself, true. But that still leaves us with the possibility of the unreal on a smaller scale. This is a sort of societal solipsism, but that in itself is not something humanity is any stranger to: Religions, especially early religions, conceived the world as a microcosm designed entirely to test humanity, a sort of Skinner box for sinners and saints.
And what if the simulation is not a total-universe simulator? Imagine one designed just for tests of humanity and human behavior. After all, if such simulators are designed by future intelligences, it stands to reason that they might be designed to test the behavior of intelligence, of consciousness, under certain frameworks?
Such a simulation would need to reproduce not an entire universe, but just enough of it to trick its inhabitants into thinking their universe was infinite. Once we grant the thought experiment of theoretical future computing power and interest, a simulation on the merely planetary scale is easier to imagine — though it would still require computing power far in advance of what we currently have.
If all that we know exists does so both entirely for and entirely within our experiencing mind, then we are all gods of our own tiny universe.
But what if the unit of simulation is not the cosmos or even the species, but the individual? How can we know what we are seeing and hearing if all of our senses are dependent on the single choke point of the brain to interpret the data? Light, sound, touch are all converted to electrical signals on their way to our experiencing self. In a way, we are all living inside a simulation either way — that is, the simulation of the world around us that our brain creates in order for us to understand and engage with it.
It is not too great a stretch of the imagination to say that those signals can be tricked and interrupted. Indeed, pop culture has already deeply engaged with this possibility, such as in the 1999 film The Matrix, in which a race of machines has enslaved the human population, connecting their brains to a simulacrum of the world to keep them docile. In fact, The Matrix is so deeply embedded into simulation theory that citing it has become almost a cliché.
“Between one in a million and 50-50,” I keep finding myself thinking when I look in the mirror.
“If you’re in a typical simulation, you shouldn’t be donating money to Africa,” Hanson says. “It doesn’t exist. You shouldn’t be saving for retirement; you’ll never retire.”
At this point in the interview, his certainty suddenly confuses me. “Unless it’s a long simulation,” I suggest.
“Right. It would have to be an unusually long one,” he says. I think, “Unusually long?” The conversation itself is starting to take on a kind of feel of unreality, which I find troubling and intriguing in roughly equal measure.
“Don’t worry about calling your mother to make sure she knows you love her, because she doesn’t exist unless you call her,” Hanson says.
The idea that what we consider to be the real world might be merely a dream has likely been around for as long as humans dreamed, but, as Gates mentioned in his email, it was explored in 1641 by the philosopher René Descartes who, in his Meditations on First Philosophy, proposed the thought experiment of the “evil genius” or “evil demon.”
Descartes supposed that a being of “utmost power and cunning” had “employed all his energies to deceive me.” As a theoretical godlike being with unlimited power of deception, Descartes’ “genius” has the power to make him “think that the sky, the air, the earth, colours, shapes, sounds and all external things are merely the delusions of dreams which he has devised to ensnare my judgement.”
“I shall consider myself as not having hands or eyes, or flesh, or blood or senses, but as falsely believing that I have all these things,” Descartes wrote.
The question he described is perhaps not as settled as Gates implied in his email. In the 1960s, Gilbert Harman, a philosopher in the United States, updated and popularized this same thought experiment. He proposed that if a mad scientist or machine removed a human brain from its body, suspended it in a jar of some liquid that kept it alive, and connected it to a computer that connected all of its input functions to a simulated reality of sufficient complexity as to trick the brain into believing it was in the real world, then there would be no way for that brain to tell the difference.
This thought experiment is key to solipsism, the philosophical outlook that takes as its fundamental axiom that nothing can be truly known for sure to be real except what is real to me. It is an outlook I find strangely comforting. If all that we know exists does so both entirely for and entirely within our experiencing mind, then we are all gods of our own tiny universe; then the distinction between that which is objectively real and that which is objectively not real becomes meaningless; the subjective experience is the singular and only component of reality.
If it’s real, that is.
“Most likely, you are in a smaller simulation around some pivotal people or events, and you probably have a good guess about who they are,” Hanson says.
For a moment, I think, horrifyingly, about Donald Trump. “Yeah,” I say, “I mean—”
“You want those people and events to continue to be interesting,” Hanson says. “And you want to continue to have an interesting relationship to them. That’s why you’re there: because they weren’t sure exactly what you would do in the simulation. That’s why they had to simulate you, and what you do matters, to some extent, to the thing they’re simulating.”
There is one part of our own history that we would love to simulate in order to understand, and that’s the moment of the primogenesis of consciousness, I think. What if some theoretical future A.I. is simulating the conditions of their birth in order to understand themselves? What if it’s not about our history at all? What if we’re not the subject of the simulation at all, but just its scenery?
I don’t say any of that, because I’m running short on time.
“Yeah,” I say weakly.