Re:John Carmack: I’m going to work on artificial general int...
by xamuel on Thursday November 14, 2019 @07:52AM
(#21534966)
As others pointed out, the NFLT only applies if the environments are uniformly distributed. In the paper, they are not uniformly distributed.
Re:John Carmack: I’m going to work on artificial general int...
by jaster on Thursday November 14, 2019 @05:50AM
(#21534425)
I missed your post and I made a similar answer.
However the idea may still be applicable if the environments votes can weighted, based on their relevance to specific domains.
(The same way optimization techniques are still useful despite the NFTL)
https://ti.arc.nasa.gov/m/profile/dhw/papers/78.pdf
I.e. across the space of all possible environments, all agents perform equally well