Header illustration by Gareth Damian Martin
After a lifetime of publishing fiction that questioned the nature of the universe, Philip K. Dick became convinced that reality was a sham.
“We are living in a computer programmed reality,” he announced to a crowd who had gathered to see the famous novelist at the Festival du livre de science fiction in Metz. In footage of the conference, which can be found on YouTube, the audience can be seen gaping on incredulously throughout. This was in 1977, three years after the event Dick referred to as 2-3-74, when he had oral surgery on a wisdom tooth and later witnessed intense visions. His intellect split open with strange ideas and new insights. He believed to have experienced the entire sum of knowledge. The visions, or hallucinations, were apparently triggered by sodium pentothal, the anesthesia used to sedate him during surgery, although the first bout didn’t occur until shortly after he had gotten home from the dentist. When a dark-haired woman, who he later would associate with a recurring character motif in his fiction, delivered his pain meds, he grew fixated on her necklace. From the pendant, he saw a beam of light shoot out. Then, he saw Christ. And geometric patterns. And an endless number of modern art paintings. Dead languages were transcribed before his eyes. He was visited by a pink laser beam that diagnosed a potentially fatal internal hernia in his son, who was rushed to the hospital. He talked to the wastebasket and it spoke back. For nights that turned into months, the cosmic entity would continue to confound him. By the end of his life, he had written 8,000 pages of largely unpublished notes dedicated to the experience.
Since this prophetic encounter with a reality-busting cosmic mind, others have joined the refrain. The majority of them have been technologists voicing similar suspicions of the soundness of the world around them. But while Dick’s obsession served as an eternal source of artistic inspiration at the tail end of an already prolific career (he had, by then, introduced several science fiction concepts that would prove depthless: alternative histories, simulated environments, AI with human consciousness), the modern myth of computer programmed reality possesses little imagination at all.
“Are You Living in a Computer Simulation?”
Most recently it was Elon Musk, the billionaire, futurist, and CEO of Tesla, to go on record as a reality skeptic. “There is one in billions chance that this is base reality,” Musk replied when asked about his views during a Q&A session at the Recode AI conference in June. His reasoning? Musk believes that a living, breathing world programmed of ones and zeros is an inevitability. Just look at videogames. In the past 40 years, games have advanced from Pong to virtual reality. By even conservative estimates, in another century or two, technology should have passed the point where game worlds and inhabitants are virtually indistinguishable from real-life people and things. “We should hope that that’s true,” he concluded, “because if civilization stops advancing then that may be due to some calamitous event that erases civilization.” Many present were laughing, almost nervously, at Musk’s remarks.
The idea that Musk was conjuring is known as the simulation hypothesis (or argument), a sort of New Age-y, tech-focused doctrine that involves elements from sci-fi, philosophy, and the singularity. The gist of the hypothesis is that nature, science, physics, and pretty much everything else is in fact an amazing software application. But how did humanity arrive to be inside a simulation in the first place? The simulation argument, as it exists in its current version, was soundly constructed by the Oxford philosopher Nick Bostrom in 2003. Though young in his career, Bostrom had already completed degrees in computational neuroscience, the philosophy of science, and philosophy, and had helped to found the transhumanist organization Humanity+, an advocate for human modification technologies, like brain implants. It is an understatement to say that he had a lot of ideas swimming around in his head. Naturally he began running these ideas by colleagues at various conferences during coffee breaks. The small talk would hatch into something big. One day, after working out at the gym (Bostrom is remarkably trim), he went to his office and typed out an argument for simulated reality. The resulting paper, “Are You Living in a Computer Simulation?” which was published towards the rear of Oxford’s Philosophical Quarterly, juxtaposed several feasible claims about future technologies into an airtight argument.
Bostrom’s thought experiment begins with a very reasonable assumption: in the future, there will be a tremendous amount of computing power. Perhaps the people in the future would use their big computers to run simulations that are advanced enough to recreate the world in intricate detail. It certainly seems possible. This part of the argument is indebted to a peer. In the paper “Simulation, Consciousness, Existence,” Hans Moravec of Carnegie Mellon’s Robotics Institute argued that, “physical science itself, through the technology of virtual reality, will provide the means to create such illusions.” But Moravec’s paper is wild and meandering. His thoughts veer into shaky territory: Plato’s cave, Cartesian skepticism, brains in vats. Bostrom avoids falling into the trap of asking: “how do we know our world is really real?” The simulation hypothesis is ultimately not an interrogation of the realness of physical reality. Of course reality is real. It’s just that the people inside the simulation exist in one part of reality, and the super-computer that is running the simulation exists in another part of reality. The theory has more to do with a boundless optimism for technology. One day, the people living inside the simulation might even decide to run their own world simulation. There could be a rabbit hole of simulations, each nestled inside the last, as long as there is enough processing power at the top to be passed down. If not, the results could be disastrous. Frying the processor would cause the world to crash. No one said that building a world simulation is without risk.
Obviously, creating such a simulation would be a Sisyphean task for normal human beings. In order to simulate a world containing billions of conscious minds, the computer would need to be big, even gigantic, perhaps the size of the moon or the Death Star. For this scenario to be tenable, humanity would have to become much more advanced than it is currently. The species would need to undergo an evolutionary leap. Bostrom refers to the people who build the computer as “posthuman.” And for humans to become post human would require a breakthrough in artificial intelligence. Right now, an algorithm’s ability to learn and think is somewhat limited. Google DeepMind, located in London, claims to have created a system that replicates neural networks, allowing a machine to learn from experience, the same way that living creatures do. So far, the system can’t quite grasp how to kill demons in Doom, yet it has mastered many simpler titles, like Space Invaders and Q*Bert, and seems to be inching towards the day where it can learn any game. Algorithms could probably learn to do much more than win at Atari.
That people are living inside simulated reality is, more or less, the only logical conclusion
The goal is an AI that can think for itself. Should an AI ever figure out how to program another AI, the world may be in for dramatic change. Some technologists, Bostrom included, believe that technology would undergo an intelligence explosion, where the intelligence of machines would accelerate with startling speed. First, the machines would build other machines. Next, the machines would start doing their own thing. Agents, or autonomous entities, such as robots who use sensors and grid overlays to move around, would have a sense of purpose. Perhaps self-driving cars would incorporate and put Uber out of business, earning the necessary means to build better self-driving cars. All the while, engineering and coding would continuously advance. Events such as these would precede a superintelligence, a kind of agent with mental capacity far exceeding that of man. This sounds problematic for the human race, but the AI could play nice, provided that humans manage to teach them to behave ethically, or wire them so that they can be powered off before going on a rampage. And thus, the humans and the machines become friends. Perhaps people will get neural implants, or have their minds uploaded into machines. Or then again, maybe not. Whatever the case, technology would have made a daring leap. A new kind of society would be achieved. Things would be great.
From there, it’s simple, really. The posthuman civilization constructs supercomputers. The computers are used to run complex simulations, universe simulators. On the screen, the world is seeded from mathematical formulas. Carbon and water mainly. Conscious beings are born in virtual space. Just microbial intelligence at first, bacteria and algae. They evolve, adapt, mutate. A long time passes. It might be possible that the system can run at double speed through the slow parts of evolution. Eventually, the sim catches up with the present tense. That people are living inside simulated reality is, more or less, the only logical conclusion.
Some futurologists balk at the idea of reality simulators. Libertarian transhumanists are the most likely to buy in to the total scenario. More conservative factions, like techno-progressivism and democratic transhumanism, probably find the prospect unrealistic. Yet the simulation argument is clever to avoid ideological conflict. Bostrom carefully carves out headroom for alternative possibilities. The paper presents three disparate options for the future, and reading through them can feel like solving an SAT question regarding dystopias. They include:
A: Everyone dies. (Human extinction could be brought on by a Skynet scenario, as in the movie Terminator, where the super-intelligent machines get here and turn out to be terrifying. While a technological apocalypse may sound like science fiction, many influential technologists, including Stephen Hawking and Bill Gates, consider the threat real enough to be taken seriously. But extinction doesn’t have to come at the hand of killer robots. It could be entirely less theatrical. Mankind could be wiped out the old-fashioned way: through nuclear winter, or when a huge meteor crashes into the planet, or from pandemic.)
B: Society evolves in such a way that future people aren’t interested in simulations. (Perhaps income inequality will be abolished in the future, or wealth is divided among mega-corporations in such a way that funding for the project becomes unlikely. It’s also possible that future people will have no tolerance for simulated pain and suffering: once AI reaches a certain bar of lifelikeness, simulations will be banned. Then again, they might never get that far. People might grow bored of digital reenactments before they are perfected, opting to use computers for other intellectual purposes, like stimulating the brain’s reward centers directly.)
C: They will in fact build the simulation. (In this case, posthumans will probably build more than one of them. Also, the chances become higher that those people are in a simulation themselves. There could be any number of simulations with conscious people living inside them. This makes the odds very slim that we are the original human beings. In all likelihood, we’re simulacra.)
Bostrom never outright claims that reality is simulated. To him, A, B, and C each have their merits. In the paper, he assigns about a 33.3 percent chance of probability to each of the outcomes. This number can vacillate. On the Simulation Argument FAQ, in response to the question “Do you really believe that we are in a computer simulation?” he writes that he thinks there is less than a 50 percent chance of it. In 2007, he told the Times that his gut feeling was more like 20 percent. His noncommittal attitude hasn’t stopped others, such as Musk, from forming their own opinions. Certainly almost everyone in the AI community has heard of Bostrom’s argument, but people seem divided on the issue. Most of the big names in AI, like Yann LeCun, who runs AI at Facebook, and Yoshua Bengio, a leading researcher in deep learning, remain unconvinced of the preconditions that need to be met for reality simulation to occur. There are simulation argument agnostics, also. Eliezer Yudkowsky, a researcher in friendly AI, founder of the internet forum Less Wrong, and author of the logically-bent fanfic Harry Potter and the Methods of Rationality, reviewed 32,142 times on fanfiction.net, told me that he “falls into the ‘has very few practical consequences’ crowd.”
the more I think about reality simulation, the more I dislike it
The simulation argument has plenty of supporters too. Of the numerous AI researchers and academia I reached out to, none thought the idea was ludicrous. Many of them actively encouraged me to write about it, but were reluctant to express themselves in interview. Even without direct comment, it’s understandable why the notion is lucrative. “There’s a certain kind of abstract thinker who’s good at math and computers and is also drawn to other similarly abstract arguments,” said Sam Frank, a journalist who covered the influx of transhumanist libertarians in Silicon Valley for Harper’s. “People who find one another online and then go to the Bay Area so they can meet similar weirdos and not feel isolated anymore are going to be attracted to rigorous, from-first-principles, unconventional arguments of all sorts.”
For instance, people who subscribe to digital physics, or the theory that everything in the universe is computational, including one professor I spoke with, are predisposed to accept the simulation story in vitro. When I asked Roman Yampolskiy, an AI safety specialist at the University of Louisville, why the view was mainly popular in computer science circles, he told me, “My guess is that people who are not that smart/educated don’t fully understand the arguments and so are not drawn to them. That is also the reason sports are a very popular topic of conversation for most people, but quantum physics is something only the smartest people talk about.”
Hubris aside, the simulation argument is surprisingly ironclad. As much as I would love to call bullshit on the whole thing, my gut reaction is hard to vindicate. Like trying to land a punch on Muhammad Ali, out-arguing Bostrom is highly unlikely. The hypothesis nimbly dodges every criticism lobbed against it. If I were to say that AI are not quote-unquote alive, a reality skeptic would counter by saying that a strong definition of consciousness is unimportant: as long as the mental processes of consciousness are emulated, a chunk of carbon-based flesh is not required. Likewise, it is a mootpoint to claim that building such an advanced computer is impossible. There may not be a large enough supply of silicon and electricity in existence to simulate the physics for the entire universe. But then again, there wouldn’t need to be. The simulation could use data compression techniques to save memory. And the computer wouldn’t have to simulate billions of people, either: Bostrom suggests that shadow people, or clones devoid of individuality, could stand in for the mass majority. The most damning criticism of reality simulation—that there is no way of testing the argument, thus deeming it beneath consideration—is also gotten around. As the argument is set up, reality is not a question of truth and falseness, but one of math and probability. It is a thought experiment—something for the people who are engineering the future to think about.
But the more I think about reality simulation, the more I dislike it. While the logic holds up to scrutiny, there is something deeply off-putting about it. As strange and irrational as it sounds, I find myself siding with Philip K. Dick’s version instead. True, Dick’s computer programmed reality was absurd, eccentric, raving. He himself dismissed it as acid flashbacks from time to time. There is evidence of pathology. Some believe Dick suffered from temporal lobe epilepsy, or TLE, a condition that predisposes one to fits of automatic writing and an inclination towards profound religious experiences. But Dick’s reality skepticism was outward-looking, a lightning bolt of ideas and discordant interpretations capable of moving in any direction. The future was constantly being reprogrammed and branching off into new nodes. Yet in the 30 or so years since Dick’s death, the idea of computer simulated reality has collapsed inward.
seeking out a new method of expression
To be frank, the simulation argument feels like watching technology perform autofellatio. Bostrom’s version of simulated reality is insular, repetitive, self-serving, logical to the point of recursion, and there is a whiff of shame and perversion about the whole process (as if AI scientists have been caught committing an act that could lead to the downfall of society). And though the argument gazes very far off into the future, it seems inexplicably shortsighted. It presents a future where there is little space for clean breaks such as scientific revolution and paradigm shifts. Without detour, current technological trends will keep moving in the present direction until AI becomes sophisticated enough to reproduce a mirror image of its distant past. That is, unless something terrible happens first. If AI doesn’t rise up against its human creators, ethics might. The argument strongly suggests that believable simulations (read: games) could be banned should society ever decide to consider AI as conscious entities. How’s that for progress?
The most offensive thing about the argument is how patronizing it is toward VR. There is an underlying assumption that the medium’s ultimate goal and purpose is to create a carbon copy of reality. Not only will reality be doubled up, but once it is effectively modeled, people will continue cloning it indefinitely, running experiments, studying themselves. This is not only a slight to the creative field of game development, but it’s antagonistic to the human imagination. While the majority of people making content for VR tend to be artists seeking out a new method of expression, the simulation argument has nothing to do with culture, the arts, or creativity of any kind. The argument treats artists, coders, players, and creative people of all types as a standing reserve towards the betterment of an idea that turns out to be the most uninventive thing in existence: a duplicate, a fake. Computer science needs better myths, or it might walk backward into replicating this one.