It’s impossible to create a fully general intelligence, i.e. one that acts intelligently in all possible universes. But we only have to make one that works in this universe, so that’s not an issue.
Well said! There might be an even stronger statement along the lines of “you can create an intelligence which is effective not just in our universe but in any universe governed by any stable local laws of physics / any fixed computable rule whatsoever”, or something like that.
The hypothetical “anti-inductive” universes where Solomonoff Induction performs worse than chance forever are very strange beasts indeed, seems to me. Imagine: Whenever you see a pattern, that makes it less likely that you’ll see the pattern again in the future, no matter what meta-level of abstraction this pattern is at. Cf. Viliam’s comment. I’m not an expert in this area but I want to go find one and ask them to tell me all about this topic :)
Well said! There might be an even stronger statement along the lines of “you can create an intelligence which is effective not just in our universe but in any universe governed by any stable local laws of physics / any fixed computable rule whatsoever”, or something like that.
I’d strengthen that to even uncomputable universes, though that requires infinite computation. The best example of an uncomputable universe is the standard model of particle physics.
It’s impossible to create a fully general intelligence, i.e. one that acts intelligently in all possible universes. But we only have to make one that works in this universe, so that’s not an issue.
Well said! There might be an even stronger statement along the lines of “you can create an intelligence which is effective not just in our universe but in any universe governed by any stable local laws of physics / any fixed computable rule whatsoever”, or something like that.
The hypothetical “anti-inductive” universes where Solomonoff Induction performs worse than chance forever are very strange beasts indeed, seems to me. Imagine: Whenever you see a pattern, that makes it less likely that you’ll see the pattern again in the future, no matter what meta-level of abstraction this pattern is at. Cf. Viliam’s comment. I’m not an expert in this area but I want to go find one and ask them to tell me all about this topic :)
I’d strengthen that to even uncomputable universes, though that requires infinite computation. The best example of an uncomputable universe is the standard model of particle physics.
Why do you say that the standard model of particle physics is uncomputable?
I think it was the constants that were uncomputable real numbers.