Thought Experiments


Do Computers Dream of Electric People?

Are we living in someone else’s fantasy?

The Chinese philosopher Zhuangzi posed this question more than two thousand years ago when he recalled waking from a dream unsure whether he was a man who dreamed he was a butterfly or a butterfly dreaming that he was a man. Today, with the advent of computers that can simulate cells, cities, and even solar systems, philosophers and scientists are asking this ancient question in a new way: Are we living in a computer simulation?

This question is more than just the premise of “The Matrix.” It’s a conjecture that lives at the intersection of humanity and technology—and though it might seem like philosophy, it spurs ambitious new questions about what computers are capable of and about the nature of reality itself. As theorists begin to think of our universe as nothing more than a vast collection of information, can we ever truly know whether our reality is as “real” as we think it is?

The philosopher Nick Bostrom, director of the Future of Humanity Institute at the University of Oxford, posed the latest iteration of this ancient question in a 2003 paper. His “simulation argument” begins with the observation that modern computers have improved at an exponential rate since their invention. If computing power continues to grow at this pace, advanced civilizations will one day be able to build titanic, densely-packed supercomputers capable of doing everything from beating the stock market to predicting the weather months or years in advance. “Post-human” programmers might even use these machines to simulate entire civilizations, vast electronic worlds that would put today’s computer games to shame.

What would it take to create this kind of simulation?

When it comes to simulating a person, scientists estimate it might take 1017 operations per second—that’s one followed by 17 zeroes—to simulate a human brain, based on the number of neurons in the brain and rate of which those neurons “talk” to each other. Assuming that simulating the sensory events a person experiences—every taste, sound, smell, touch and sight that is coded in our neurons—takes about 100 million bits per second, and that approximately 100 billion humans have lived on Earth to date, Bostrom estimates it might take 1036 calculations in total to create a simulation of the whole of human history that is indistinguishable from reality.

That’s just to simulate the parts of the universe that humans can sense. What about the microscopic structure of the Earth’s interior or the subtle features of distant stars? These little details could be safely omitted until a simulated person needed to observe them. In addition, to save computing power, maybe not every person in a simulation would be fully simulated. Perhaps some of the characters in the simulation would be “zombies or ‘shadow-people’—humans simulated at a level sufficient for the fully simulated people to not notice anything suspicious,” Bostrom writes in his paper.

So how close are we to achieving this dream (or nightmare)? Today’s most powerful supercomputers are capable of operating at roughly 10 petaflops per second—that is, 1016 calculations per second. A planet-sized computer based on current electronics might carry out 1042 operations per second. Bostrom also notes that quantum physicist Seth Lloyd of MIT has calculated that a 1-kilogram “ultimate laptop” that operates at the known limits of physics might be capable of 5 × 1050 operations per second. So, the planet-sized computer might be able to simulate all of human history in a millionth of a second; the ultimate laptop, a hundredth of a billionth of a second.

Given that fully simulating every person who has ever lived might only take a tiny fraction of an advanced civilization’s resources, Bostrom reasons that the number of computer-generated minds buzzing away inside simulations could vastly outnumber the total sum of real minds that have ever lived. If that is true, the odds are that we are simulated, not real. It may even be possible that our simulators are themselves simulated, and their simulators are simulated, and so on. “Reality may thus contain many levels,” Bostrom says.

This does not prove that we live in a simulation, Bostrom emphasizes. There are a number of caveats that could stop this bizarre future before it starts. One glum possibility is that civilizations might very well go extinct or collapse—say, by annihilating themselves in a nuclear war—before they can develop supercomputers of such immense power. Another thought is that civilizations simply have no desire to commit the vast resources needed to create supercomputers. Or perhaps advanced civilizations might not indulge in such simulations—maybe they would be ethically opposed to simulating minds and their suffering, or they might prefer to entertain themselves with machines that directly stimulate their brain’s pleasure centers. “Personally, I assign less than 50 percent probability to the simulation hypothesis—rather something like in the 20 percent region, perhaps, maybe,” Bostrom writes, although he describes this as a gut feeling rather than part of his logical argument.

Unless the simulators decide to make themselves known, there may be no way to prove or disprove the simulation argument. Some have suggested looking for “glitches” in the simulation, but such glitches would be more plausibly explained as hallucinations, visual illusions, fraud or self-deception. Even if errors did pop up, a smart simulator could simply wipe any memory of the anomaly from our simulated brains.

If we are living in a computer simulation, how should we live our lives? “The simulation hypothesis currently does not seem to have any radical implications for how one should live,” Bostrom said. Still, “it helps to shed light on, among other things, the prospects of our species.”

Also, thinking of the universe as a computer may actually be a helpful approach in science. “You can start thinking about what kind of computer it is, what kind of operations can it do, what kinds of problems can it solve,” said theoretical computer scientist Scott Aaronson at MIT. “That’s an extraordinarily fruitful way of thinking about the universe that has led to the whole field of quantum computers—devices based on the quantum physics that explains how the fundamental building blocks of the universe behave.”

We may never know whether we are living in someone else’s fantasy; whether we’re the man or the butterfly. But if we do one day develop supercomputers capable of simulating minds and universes, perhaps our creations will be able to answer the question for us.

Tell us what you think on Twitter, Facebook, or email.

Charles Choi

    Charles Q. Choi has written for Scientific American, The New York Times, Wired, Science and Nature, among others. In his spare time, he has traveled to all seven continents, including scaling the side of an iceberg in Antarctica, investigating mummies from Siberia, snorkeling in the Galapagos, climbing Mt. Kilimanjaro, camping in the Outback, avoiding thieves near Shaolin Temple and hunting for mammoth DNA in Yukon.