Originally published January 15, 2002 on KurzweilAI.net.

A Beautiful Mind — a fascinating exploration of the nexus of genius and madness of eminent mathematician and Nobel prize in economics winner Dr. John Nash –is my nomination for the best accelerating-intelligence-related movie of the millennium (so far). Intellectually engaging plot, humor, great acting.

Best part: an explanation of the “Nash equilibrium”–a key concept in modern game theory–in a bar pickup scene. (It helps if you read the book first.)

Yes, cracking enemy codes in newspaper and magazine ads and being chased by Russians in a CIA getaway car is fictional. But Nash, who transformed von Neumann’s game theory with his PhD dissertation in 1950, was indeed obsessed with finding obscure patterns–including those from ETs who were, he says, “communicating with him” through the newspaper.

And who’s to say definitively they weren’t? “The ideas I had about supernatural beings came to me the same way my mathematical ideas did,” said Nash. “So I took them seriously.”

Reality vs. virtuality is also at the core of the mind-bending Vanilla Sky, number two on my list. The central character, David Ames, played by Tom Cruise, is disfigured in a car crash, has his face restored by plastic surgery, and finds himself being interrogated by a police psychologist for his alleged involvement in the bizarre murder of his girl friend, who morphs into another woman.

Like Nash, Ames seems to be involved in a bizarre conspiracy. But in a dramatic twist, it turns out he actually died after being in coma following the crash and was put in cryonic storage, having chosen to have a VR experience while in suspension — presumably by having his brain downloaded prior to the accident.

The film joins the growing VR genre, including Brainstorm, The Lawnmower Man, Strange Days, The Matrix, eXistenZ, The Thirteenth Floor, and Being John Malkevich.

“It’s the first film that deals intelligently with the issues raised by cryonics,” as cryonics pioneer and Alcor co-founder Mike Darwin was quoted on the Extropians list. “It is the first movie that captures the imagery created by cryonics technology (Alcor and Trans Time style) as filtered through the media. It was very, very strange to see the world I have participated in creating in microcosm writ large, and now clearly a part of The Culture.”

An even stranger movie is my number 3, Richard Linklater’s Waking Life. Using rotoscoping to convert live action into impressionistic, quasi-animated, dreamy sequences, this head-trippy film portrays the adventures of a twenty-something guy floating through a series of multilayered, free-associational philosophical and visionary musings about the nature of consciousness, existentialism, and lucid dreaming, while trying to figure out if he’s awake, dreaming, or dead.

He’s given one hint: if you turn on a light switch and nothing changes, you’re dreaming. He wakes up, turns on a light switch, and nothing changes.

The highlight of the film for me is a hypercaffeinated Kurzweilian character who expounds on the exponential growth of life and intelligence on the planet over the past four billion years.

Are you living in a computer simulation?

So what’s real and what’s Memorex? In his essay, “Are You Living In a Computer Simulation,” Dr. Nick Bostrom claims there’s no way to know whether or not you’re living in an “ancestor- simulation” (simulation of us by our descendents) created by some future posthuman civilization to amuse or educate themselves by reconstructing the past. [1]

Of course, we currently don’t have the computer power or software to create conscious minds. But as Ray Kurzweil has said in The Age of Spiritual Machines, that might be achievable as early as 2020, when a massively parallel, neural-net personal computer will have the calculation capacity of the human brain, estimated at 2 x 10¹⁶ calculations per second — or even by 2010, using a supercomputer. (It might take another ten years to develop the needed software, he adds.) Based on his “Law of Accelerating Returns” (double exponential growth), in just a few centuries, computer capacity would be beyond our imagination.

“Such a mature stage of technological development will make it possible to convert planets and other astronomical resources into enormously powerful computers,” says Bostrom.

To be a true simulation, you would need to simulate the real world from the universe down to the quantum level, which would require mind-boggingly massive real-time computations. But, he argues, “to get a realistic simulation of human experience, much less is needed — only whatever is required to ensure that the simulated humans, interacting in normal human ways with their simulated environment, don’t notice any irregularities.”

Bostrom says the “main computational cost consists in simulating organic brains down to the neuronal or sub-neuronal level.” But that assumes a neuromorphic design (based on the structure and operation of the brain) is required. Carver Mead, who pioneered neuromorphic design, tells me that he now believes non-neuromorphic algorithms are more efficient in simulating brain functions.

Perhaps it would also be more efficient to always generate ad hoc simulations that are based on the current attention state of each individual, rather than continuous simulation of the entire world. We could probably also achieve savings in storage requirements by representing objects in nature with fractals.

For more computational efficiency, you could also omit the structure of the inside of the Earth and use compressed representations for distant astronomical objects, says Bostrom. “The posthuman simulator would have enough computing power to keep track of the detailed belief-states in all human brains at all times. Thus, when it saw that a human was about to make an observation of the microscopic world, it could fill in sufficient detail in the simulation in the appropriate domain on an as-needed basis. Should any error occur, the director could easily edit the states of any brains that have become aware of an anomaly before it spoils the simulation. Alternatively, the director can skip back a few seconds and rerun the simulation in a way that avoids the problem.”

To do such an ancestor simulation of the entire mental human history to date would require ~10³⁷ operations [2], polymath computer programmer Robert Bradbury calculates. Or if we accept Stuart Hameroff’s speculation that “brain processes relevant to consciousness extend downward within neurons to the level of cytoskeletal microtubules,” 10²⁷ operations per second would be required to simulate a single person’s mental history, or ~10⁴⁷ operations for the entire human history.

But let’s stay with the lower figure for now. Bradbury says it could be achieved with a “Matrioshka Brain” [MB, a megascale computer constructed as a series of shells around a star using nanoscale components], providing roughly 10⁴² operations per second. With some of the more advanced nanoscale computational architectures and cooling methods, you might be able to push this up by perhaps as much as 10¹⁰ operations per second.”

Assuming the lower figure, an MB could run the entire ancestor- simulation of human history in less than 10^-5 second [3]. Or else it could run 100,000 instantiations of virtual mental worlds simultaneously (parallel worlds). Interaction between these parallel realities would allow for some interesting anomalous experiences, including “time travel” (the interaction would add significant computational load and limit the number of instantiations, of course).

But the MB design assumes known nanotech design principles, “probably far from optimal,” says Bostrom. “If we could create quantum computers, or learn to build computers of nuclear matter or plasma, we could push closer to the theoretical limits. Posthuman civilizations would have enough computing power to run hugely many ancestor-simulations even while using only a tiny fraction of their resources for that purpose,” he says.

At the extreme, Seth Lloyd of MIT has calculated the conceivable limits of computing power of a black hole computer: a maximum processing speed of about 10⁵¹ operations/sec (equivalent to 10³³ IBM Blue Gene supercomputers) for a 1-kg black hole.

To take it even further out, “It may be possible for simulated civilizations to become posthuman,” Bostrom adds. “They may then run their own ancestor-simulations” using powerful virtual machines they build in their simulated universe (these virtual machines themselves could be stacked in multiple levels.)

What’s more, “We would have to suspect that the posthumans running our simulation are themselves simulated beings; and their creators, in turn, may also be simulated beings. Reality may thus contain many levels. Even if it is necessary for the hierarchy to bottom out at some stage — the metaphysical status of this claim is somewhat obscure — there may be room for a large number of levels of reality, and the number could be increasing over time.”

Of course, all this assumes the human species doesn’t go extinct first by some massive nuclear or biological catastrophe or global ecophagy (the “gray goo” scenario) in which out-of-control nanotech replicators wipe out all life on Earth–although Robert Freitas does offer some possible solutions.

Another possibility is that simulating even a single posthuman civilization might be too expensive, says Bostrom. ” If so, then we should expect our simulation to be terminated when we are about to reach the posthuman stage.” Talk about spoil sports!

How to live in a simulation

Are you still with me? OK, let’s assume our posthuman successors have somehow made it to the future and they’re running a simulation, starring … us. What should we do–and would we have any control?

“The only way to influence the real world is to somehow influence whoever is observing this simulation,” points out Robin Hanson in his Bostrom-inspired essay, “How To Live In A Simulation.”

Since such simulations would be very costly, they would probably be limited in time and space. To make the simulators less likely to drop you from their simulation or end it, you should be entertaining–“funny, outrageous, violent, sexy, strange, pathetic, heroic … in a word ‘dramatic.’ And since the simulators may want to play famous people, you should “keep famous people happy, or at least interested… and try to stay personally interesting to the famous people around you.” Sounds like a typical Hollywood cocktail party.

The simulators might also want to play God, “punishing and rewarding people in the simulation based on how they lived their lives.” In that case, “you’ll have to figure out the common features of morality in descendants who are willing to play God.”

Getting a clue

So unless you are a God, how would you know you were living in a simulation? Well, there might be glitches or discontinuities. Ames figured out he was in a simulation when he encountered his VR tech support person, who was then able to modify the scenario in real time. In the movie Pleasantville, two contemporary teenagers arrive back in the 50s and influence it to transform into the 60s. In Fred Pohl’s “Tunnel Under the World” scifi story, the hero discovers that the world is just empty space when he turns over a boat in his basement and figures out he’s in a simulation created on a subatomic scale. That could ruin your day.

Bradbury suggested three other examples: Theodore Sturgeon’s scifi “Yesterday Was Monday” (an auto mechanic wakes up and finds tiny blue men preparing the “set” for the next “act” the next day), the movie Dark City (soulless entities known as The Strangers freeze time every night to rearrange the skyline and warp every resident’s mind to find out what makes us “tick”) and the movie The Truman Show (a man’s entire life is scripted as programming for a 24-hour television network). Innovator Mike Lorrey also suggested Stephen King’s movie “The Langoliers” (ten passengers on a plane awaken to find everyone else in the world has disappeared and figure out how to get back into reality).

Another clue might be anomalous experiences, which might imply defective algorithms or hardware. For example, telepathy might imply “leaks” between data sets and UFOs, alien abductions and “miracles” might imply deliberate attempts to mess with our minds to test theories or for amusement. [4]

Bostrom added these tips: “You’d want to know what kind of simulations the simulators would be most interested in running, and the cost of these different kinds of simulations. Then the simulation hypothesis would predict that we should find ourselves in a world that ranks high on interest and low on cost. Right now, we don’t have much evidence to go on in either these regards. (Remember that what we take as the fundamental physics may be largely merely apparent in a simulation, so we can’t necessarily infer the real cost of running an adequate simulation of our experiences from the computational complexity of what we take to be physical laws.)

“Then there are obvious tipoffs, e.g., the simulators could tell us that we are in a simulation or even extract us from it into their own level of reality, so the hypothesis is clearly not conceptually unverifiable.”

Tools for creating such future simulations are evolving rapidly. Already, with films like Final Fantasy, the best video games for PlayStation and Xbox, and the most powerful military simulators, the boundaries between real and virtual are dissolving as simulation representation approaches the limits of the “real” (whatever that means) world.

In just seven years, in 2009, we’ll have “routine, full-immersion, visual-auditory, virtual-reality shared environments with images written directly to our retinas from our eyeglasses and contact lenses,” says Ray Kurzweil. “By 2029, ‘experience beamers’ will beam their entire flow of sensory experiences and feelings onto the Web the way people now beam their images from their web cams.”

At that point, synthesized simulacra and the real world may be indistinguishable, as will perhaps cyborgs and people. Telling the difference will become the challenge of the future — and the subject of future films.

Amara D. Angelica is editor of KurzweilAI.net

Footnotes:

[1] Of course, reading this could be dangerous if “They” don’t want anyone to know (à la the movie Conspiracy Theory) and decide to terminate your simulation. You have been warned :)

[2] The ~10³⁷ figure is “based on ~10¹¹ humans * ~50 years/human * 31*10⁶ secs/year * 10¹⁷ operations in each human brain per second,” said Bradbury. “The total number of humans and their life times are fuzzy because population estimates before the last couple of hundred years are all guesses (not on my part, but on the part of the people who have published data on the topic). We also know that the population may have gone through one or more bottlenecks so you can’t assume a steadily increasing population.”

[3] “One critical thing that must be kept in mind is network delays,” said Bradbury. “You have to keep the nanocomputers separated by not-so-insignificant amounts of space (meters to many kilometers) because of the heat they radiate. So while you have the processing power to run an ancestor simulation in 10^-5 aggregate CPU seconds, you may not be able to actually run the simulation in 10^-5 “real” seconds because of the time that individual nodes will have to wait to get information about something that happened in another part of the simulation.

“Obviously you could optimize things by putting ‘minds’ that require close connectivity on the same ‘node’ or adjacent nodes of the net. So how fast in real time you could run an ancestor simulation depends in large part how interconnected individuals within the simulation are. The more interconnected they become, the slower the simulation will run in real time.

“In very densely interconnected societies, where everyone may need to relate to everyone else, a Jupiter Brain [a posthuman being of extremely high computational power and size] may be an optimal simulation architecture. However, for loosely connected societies, such as that currently found on Earth, a Matrioshka Brain seems more likely to be an optimal simulation architecture because it provides greater computational capacity and allows closely interacting elements in the simulation to be located on relatively nearby network nodes.”