Re: What are useful for a phd?

From: Eliezer S. Yudkowsky (
Date: Tue Dec 05 2006 - 15:07:33 MST

Hector Zenil wrote:
> The complexity theory you are referring to is called "computability
> theory".

Computability theory is for Turing degrees; and for what used to be
called "recursive", "recursively enumerability" and are now starting to
be called "computable" and "computably enumerable" functions.
"Computational complexity" is the O(n log n) stuff and as distinguished
from "complexity studies" a la Santa Fe. Incidentally, I've never heard
of anything remotely AI-relevant coming out of complexity studies. In
my humble opinion. Complexity studies are for physics-type phenomena,
not phenomena subject to optimization processes, such as biological
phenomena and cognitive phenomena.

To the lists being given so far, and if you are genuinely interested in
the F part of FAI, I add:

Evolutionary biology with math;
Evolutionary psychology;
Heuristics and biases;
Decision theory as taught by analytic philosophers;
and at least one course on functional neuroanatomy and the pathology of
brain-damage disorders.

But mostly it would depend on what you want to do. Invent the whole
field of Friendly AI from scratch? Don't expect that you can just get a
Ph.D. and then go out and do that. If you're taking a narrower approach
and you have some specific issue you want to make progress on, then
stating the specific issue could let me tailor down the recommendations
a bit.

Eliezer S. Yudkowsky                
Research Fellow, Singularity Institute for Artificial Intelligence

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:57 MDT