The idea that reality as we experience it may be a technologically
based simulation is a staple of science fiction, showing up in films
such as The Matrix, Total Recall
and Vanilla Sky.
It also has been of some interest to philosophers, who raise scenarios
involving brains in vats and so forth as a way of contemplating the
nature of knowledge.
In the past couple of years, though, the idea has gained some
credence as a plausible or likely description of how things actually
are. Philosopher Nick Bostrom of Oxford University has presented a simulation argument
which holds that there is a significant chance we do live in a
simulation. His reasoning runs as follows: A technologically advanced
civilization (such as humanity may become) would have enormous
computing power, which could be used to create numerous simulations
(complete with consciousness) of the civilization's own ancestors.
Thus, there are three possibilities (which, in the absence of further
information, Bostrom sees as equally likely):
One, such advanced civilizations almost never come into existence
-- which suggests humanity is likely to go extinct before developing
the requisite advanced technologies. Two, advanced civilizations have
almost zero interest in running simulations of their ancestors. Or
three, the overwhelming majority of people with experiences like ours
are simulated -- in which case we almost certainly do live in a
Drawing upon Bostrom's ideas, Robin Hanson, an economist at George Mason University, wrote a paper
on "How to Live in a Simulation." Hanson's guidelines are not so much
an ethical theory as a way to minimize the possibility the simulation's
controllers will get bored and pull the plug. Among other things, he
suggests that people should "care less about others," "live more for
today," and "keep the famous people around you happy and interested in
But are there really any substantial grounds for thinking we live
in a simulation? A look at the assumptions and ambiguities built into
the simulation argument suggests otherwise.
There is, to begin with, the question of whether consciousness can
be created in a computer simulation. The slowness and difficulty of
progress in artificial intelligence gives some reason for skepticism on
this point. Nonetheless, as Bostrom points out, his argument requires
no assumption about timescale; computer consciousness might be possible
100,000 years from now, if not in 50 or 100 years. This seems
reasonable enough, although the possibility that computer consciousness
may be impossible cannot be ruled out. If it is impossible, the
simulation argument fails.
Let's assume, though, that our distant descendants can create such
conscious simulations. The next question is what they would do with
this power. Bostrom assumes that at least some of them would create
large numbers of simulations of their ancestors. He notes that even if
such things are illegal or frowned upon, all it would require is one or
a few wealthy individuals with access to vast computer power and an
interest in ancestor-simulations. Such simulations could be done using
only a tiny fraction of the enormous computing resources that the
advanced civilization is presumed to have.
But there's a problem here. Consider what computer simulations are
used for today, particularly the vast assortment of games involving
dragons, aliens and so on. Ancestor-simulation is low on the list. Even
if our descendants have greater interest in ancestor-simulation than
humans do today, there remains the likelihood that a far larger set of
conscious simulations will be created that are not about realistic
depictions of ancestors.
Thus, while simulated people may outnumber real people, people in
nonrealistic simulated worlds should outnumber those in realistic
simulated worlds. Notwithstanding accounts of alien abductions and the
like, we appear to live in a realistic world -- which should puzzle us,
if the simulation argument is true. We do not appear to be typical
observers, yet the idea that we are typical is what led us to think
we're simulated (since it is assumed that only a small fraction of
people are real).
The simulation argument, in this respect, is similar to the doomsday argument
(also popularized by Bostrom), which claims it is improbable we would
find ourselves alive today if there will be numerous humans alive in
the future. In the doomsday argument, it is not clear if only humans
belong in our reference class or also aliens, sentient animals and
posthuman robots. In the simulation argument, it is not clear what kind
of simulated beings we should count ourselves among.
Then again, if the simulation argument is true, we have little
reason to think that the real universe (where the programmers live) is
anything like our own universe. It is possible that the programmers are
giant octopuses, or that their universe has totally different laws of
physics. This would solve the problem of the unexpected realism of our
simulation. But it also undercuts our ability to form any assumptions
whatsoever based on our own experience. The simulation argument
requires such assumptions (for example, the assumption that our
descendants will exist and want to simulate their ancestors' lives).
Finally, it is worth asking whether, if the simulation argument
were true, the programmers would allow their creations to engage in
conscious reflection about their true situation. One would think that
speculations about simulated universes would be edited out, and that
articles such as the one you're reading would never be allowed. The
fact that you are reading this proves that ? GLITCH.
Note to film buffs: If you watch the 2001 movie
look for the scene early on where actor Jason Leigh walks down a
building's steps and meets Tom Cruise in a waiting car. The scene was
filmed on the stoop of the real-life building where I live.