My current view on this is that you can shift to the language of higher-order logic, with the ordinals for the order of logic that are provable in some proof system. Not sure you can do any better of that.
Does that help? IIRC there’s a level in the computability hierarchy for every countable ordinal number, and no way of having a probability distribution that assigns nonzero mass to every countable ordinal.
Earlier you said that Solomonoff induction is enough to beat any computable human, even if the universe is uncomputable. What made you change your mind?
Has he changed his mind? What I see here is an indication that you can extend Solomonoff induction to have uncomputable possibilities in your prior, but how does that imply that a computable human can do better than Solomonoff induction?
My current view on this is that you can shift to the language of higher-order logic, with the ordinals for the order of logic that are provable in some proof system. Not sure you can do any better of that.
Does that help? IIRC there’s a level in the computability hierarchy for every countable ordinal number, and no way of having a probability distribution that assigns nonzero mass to every countable ordinal.
Do you know any counterpart to the Solomonoff distribution for universes based on higher-order logic?
I’d very much like you to write more on such topics because you seem to have really good math intuition, we could use that.
That I didn’t invent? No.
Earlier you said that Solomonoff induction is enough to beat any computable human, even if the universe is uncomputable. What made you change your mind?
Has he changed his mind? What I see here is an indication that you can extend Solomonoff induction to have uncomputable possibilities in your prior, but how does that imply that a computable human can do better than Solomonoff induction?