You're looking at Less Wrong's discussion board. This includes all posts, including those that haven't been promoted to the front page yet. For more information, see About Less Wrong.

V_V comments on Versions of AIXI can be arbitrarily stupid - Less Wrong Discussion

15 Post author: Stuart_Armstrong 10 August 2015 01:23PM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Article Navigation

Comments (59)
Tags: ai optimality aixi

Comments (59)

You are viewing a single comment's thread. Show more comments above.

Comment author: V_V 11 August 2015 06:25:45PM -1 points [-]

my instincts are that while this problem might exist, you would only actually run into it if using a language specifically designed to create this problem.

I think it's actually worse. If I understand correctly, corollary 14 implies that for any choice of the programming language, there exist some mixtures of environments which exhibit that problem. This means that if the environment is chosen adversarially, even by a computable adversary, AIXI is screwed.

Comment author: skeptical_lurker 11 August 2015 07:19:17PM 0 points [-]

Hmm, well you can't choose the laws of physics adversarially, so I think this would only be a problem in a pure virtual environment.

Comment author: V_V 11 August 2015 07:36:24PM -1 points [-]

The laws of physics may allow for adversaries that try to manipulate you.

Powered by Reddit Powered by Reddit