Monday, November 13, 2017

Maximizing options as the basis of memory-less intelligence

There seems to be some big news in A.I. and cosmology. To give an example of how far-reaching this idea is, view walking upright and the ability to use our hands as something that maximizes our options rather than something that gives us more power in any other sense. This simple algorithm can successfully play games without any training at all, other than defining "maximize future options" for a given set of rules.

I am far from certain his general view is correct (or even expressed clearly enough to criticize. I still can't view it as anymore than a method of problem solving when his claims are much greater than that, but I can't exactly understand his claims. It sounds like he is saying things appear intelligent because nature somehow keeps options open.


Ted Talk
https://www.ted.com/talks/alex_wissner_gross_a_new_equation_for_intelligence

General idea:
https://physics.aps.org/articles/v6/46

How it's good at playing atari without training:
http://entropicai.blogspot.com/2017/06/solved-atari-games.html#more

On freedom in society making us more powerful:
https://www.edge.org/response-detail/26181

Basic physics of causal entropy and intelligence
http://www.alexwg.org/publications/PhysRevLett_110-168702.pdf

How it can predict the cosmological constant by following the anthropic principle:
http://www.slac.stanford.edu/cgi-wrap/getdoc/slac-pub-12353.pdf


No comments:

Post a Comment