Le 02-févr.-06, à 17:31, Norman Samish a écrit :
My conjecture is that a perfect simulation by a limited-resource AI would not be possible. If this is correct, then self-aware simulations that are perpetually unaware that they are simulations would not be possible.
This could be a reasonable conjecture. I have explain on the list that if we are a simulation then indeed after a finite time we could have strong evidence that this is the case, for exemple by discoveries of discrepancies between the "comp-physics" and the "observed physics".
Humans have not made the discovery that they are simulations, therefore the most probable situation is that we are not simulations.
This argument is not valid. The reason is that if we could be "correct" simulation (if that exists), then that would remain essentially undecidable.
Then I could argue the premise is false. Violation of bell's inequalities could be taken as an argument that we are in a simulation (indeed in the infinity of simulation already "present" in the "mathematical running" of a universal dovetailer, or arithmetical truth.