I asked Marvin Minsky, a legendary founder of artificial intelligence, to distinguish among three kinds of simulations: (i) brains in vats, (ii) universal simulation as pure software and (iii) universal simulation as real physical stuff.
"It would be very hard to distinguish among those," Minsky said, "unless the programmer has made some slips — if you notice that some laws of physics aren't quite right, if you find rounding-off errors, you might sense some of the grain of the computer showing through."
If that were the case, he says, it would mean that the universe is easier to understand than scientists had imagined, and that they might even find ways to change it.
The thought that this level of reality might not be ultimate reality can be unsettling, but not to Minsky: "Wouldn't it be nice to know that we are part of a larger reality?" [Incredible Technology: How Future Space Missions May Hunt for Alien Planets ]
For a reality check, I visited Martin Rees, U.K. Astronomer Royal, a bold visionary and hard-nosed realist.
"Well, it's a bit flaky, but a fascinating idea," he said. "The real question is what are the limits of computing powers."
Astronomers are already doing simulations of parts of universes. "We can't do experiments on stars and galaxies," Rees explained, "but we can have a virtual universe in our computer, and calculate what happens if you crash galaxies together, evolve stars, etc. So, because we can simulate some cosmic features in a gross sense, we have to ask, 'As computers become vastly more powerful, what more could we simulate?'
"It's not crazy to believe that some time in the far future," he said, "there could be computers which could simulate a fairly large fraction of a world."
A prime assumption of all simulation theories is that consciousness — the inner sense of awareness, like the sound of Gershwin or the smell of garlic — can be simulated; in other words, that a replication of the complete physical states of the brain will yield, ipso facto, the complete mental states of the mind. (This direct correspondence usually assumes, unknowingly, the veracity of what's known in philosophy of mind as "identity theory," one among many competing theories seeking to solve the intractable "mind-body problem".) Such a brain-only mechanism to account for consciousness, required for whole-world simulations and promulgated by physicalists, is to me not obvious.
I asked Rees whether human-level consciousness and self-consciousness can be simulated.
"That may be the kind of question that would demand a superhuman intelligence to answer," which, he adds, "could be forever beyond our capacity."
Physicist Paul Davies has a different take. He uses simulation theory to tease out possible contradictions in the multiple universe (multiverse) theory, which is his countercultural challenge to today's mainstream cosmology.
"If you take seriously the theory of all possible universes, including all possible variations," Davies said, "at least some of them must have intelligent civilizations with enough computing power to simulate entire fake worlds. Simulated universes are much cheaper to make than the real thing, and so the number of fake universes would proliferate and vastly outnumber the real ones. And assuming we're just typical observers, then we're overwhelmingly likely to find ourselves in a fake universe, not a real one."
So far it's the normal argument.
Then Davies makes his move. He claims that because the theoretical existence of multiple universes is based on the laws of physics in our universe, if this universe is simulated, then its laws of physics are also simulated, which would mean that this universe's physics is a fake. Therefore, Davies reasoned, "We cannot use the argument that the physics in our universe leads to multiple universes, because it also leads to a fake universe with fake physics." That undermines the whole argument that fundamental physics generates multiple universes, because the reasoning collapses in circularity.
Davies concluded, "While multiple universes seem almost inevitable given our understanding of the Big Bang, using them to explain all existence is a dangerous, slippery slope, leading to apparently absurd conclusions."
Five premises to the simulation argument
I find five premises to the simulation argument: (i) Other intelligent civilizations exist; (ii) their technologies grow exponentially; (iii) they do not all go extinct; (iv) there is no universal ban or barrier for running simulations; and (v) consciousness can be simulated.
If these five premises are true, I agree, humanity is likely living in a simulation. The logic seems sound, which means that if you don't accept (or don't want to accept) the conclusion, then you must reject at least one of the premises.
Which to reject? Other intelligent civilizations? Exponential growth of technology?
Not all civilizations going extinct? No simulations ban or barrier? Consciousness simulated?
Whichever you choose, it must apply always, everywhere. For all time. In all universes. No exceptions.
That, to me, makes no sense.
Would the simulation argument relate to theism, the existence of God? Not necessarily.
Bostrum said, "the simulation hypothesis is not an alternative to theism or atheism. It could be a version of either — it's independent of whether God exists." While the simulation argument is "not an attempt to refute theism," he said, it would "imply a weaker form of a creation hypothesis," because the creator-simulators "would have some of the attributes we traditionally associate with God in the sense that they would have created our world."
They would be superintelligent, but they "wouldn't need unlimited or infinite minds." They could "intervene in the world, our experiential world, by manipulating the simulation. So they would have some of the capabilities of omnipotence in the sense that they could change anything they wanted about our world."
So even if this universe looks like it was created, neither scientists nor philosophers nor theologians could easily distinguish between the traditional creator God and hyper-advanced creator-simulators.
But that leads to the old regress game and the question of who created the (weaker) creator-simulators.
At some point, the chain of causation must end — although even this, some would dispute.
Personally, I do not think humanity is living in a whole-world simulation. But because the simulation argument seems to work, what it seems to do is to uncover deep discrepancies, or fundamental flaws, in how people think about deep reality — about this universe, multiple universes, consciousness, and even inferences for and against theism.