PAPER: Theory of Universal AI based on Algorithmic Complexity

From: James Rogers (
Date: Sun Apr 15 2001 - 17:14:33 MDT


I've been working on developing a mathematically thorough top-down model of
AI for several years now, working from a substantially different perspective
than a lot of AI work. I've been working on distilling the complex topic
down into a paper on a very part-time basis for a couple years now.
Unfortunately, as I approach completion it seems that a german mathematician
beat me to the punch by a matter of months. My own implementation work has
based on a nearly identical mathematical model of AI.

Since his paper overlaps about 90% of mine, I would refer all the readers on
this list to his paper. Hell, about 2/3 of his mathematics references match
mine. I find it interesting that not only did he arrive to the same
mathematical model, but that he appears to have developed some of the same
biases towards a lot of AI research that I have as well as a result of the

Where we differ: In content, we don't -- he is right on the money. We
differ only in that I have spent substantially more time exploring certain
specific areas and implementation issues in general i.e. My coverage detail
is broader, though his is by no means narrow. He also makes some
conjectures with strong support, but is apparently unaware that important
proofs of his conjectures have been published in the last several years.

I strongly encourage people to read at least the introduction (the paper
itself is approx. 65 pages long) to get an idea of the mathematical
concepts, theory, and implications. The paper itself can be very terse and
mathematical at times but is digestible, particularly if you jump over the
formalities. I am very interested in hearing this list's feedback on the

I'm only annoyed because I spent some of my limited time writing when I
could have been coding... :^)


-James Rogers

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:36 MDT