驗證我們是否處於模擬現實中

Everything that we perceive of meaning and value, is related to what we perceive. The question of whether our perceived reality is "really real" as opposed to a dream, hallucination, fabrication or (in modern times) simulation is an ancient and vexing one.

Virtual Reality and Artificial Intelligence, dramatically extrapolated, suggest paths by which we might be denizens of a simulated world.

In what we might call "natural consciousness, virtual reality" (NCVR), a (presumably) biologically based consciousness would be fooled, via its sensory inputs, into believing it exists in a very different reality than its physical body. This is the conceit behind, for example, The Matrix movie series and the "brain in a vat" set of philosophical explorations.

A different mode, "artificial consciousness, simulated reality" (ACSR) considers artificially intelligent and conscious computer programs that experience a fully fabricated reality – what might be true (under certain philosophical assumptions) if a non-playing character in a video game were sufficiently complex as to have an internal subjective experience. (This would also describe the "Agents" in The Matrix.)

While this is (literally!) the stuff of science fiction, there are cogent arguments put forth in the literature, for example by Bostrom, that we are likely to be in a simulated reality. In a nutshell, if such simulations are possible, it is likely that somewhere in spacetime they will be performed, and arguably give rise to many, many more simulated beings than does the "real world." Thus if we assume just that we are subjectively self-aware beings, we are more likely to be one of the simulated ones than a "real" one.

This argument has recently been in the news when Elon Musk asserted that “There’s a one in billions chance [we’re in] base reality... I think it’s one in billions."

While primarily a topic of late-night conversations after a few drinks, it does not seem impossible for this to be the subject of some rigorous research, and careful investigation might well lead to interesting insights into, for example, theories of consciousness, artificial intelligence, the interpretations of quantum mechanics, the anthropic principle, etc. On the other hand, investigating a problem at the intersection of four inscrutable and intractable problems may just be the road to madness. Which will it be?

The problem is to come up with a rationale, or principle to verify if we are living in a simulation or not, and to observationally/experimentally distinguish some mode of ACSR or NCVR simulation from "physical reality" as generally construed. The observable need not be generic to ACSR or NCVR scenarios (or both), just be able to address at least some versions in a nontrivial way.

By Anthony, originally posted at Metaculus - a community focused on probabilistic mapping of the future.