From: Emil Gilliam (firstname.lastname@example.org)
Date: Wed Jul 17 2002 - 22:34:42 MDT
I think I understand the argument of why SIAI wants to build an "infant"
(so to speak) rather than a "genome" for its seed AI.
As a separate issue, however, I am really nervous about how so few genes
(some fraction of the 30,000 to 120,000 genes that are believed to exist
in the human genome, by various estimates) can encode wiring algorithms
for the many purported built-in modules of the brain that are supposed
to exist under the Integrated Causal Model. Can anyone give, or point me
to, a really satisfying hypothesis for this?
It is not sufficient to point out the number of megabytes of data
encoded in the raw DNA comprising these genes, as this is meaningless in
and of itself. It is not enough merely to point out the notorious
complexity of protein folding, at least without some good argument about
"how much" of this complexity gives rise to the functional complexity we
see in the mind (or any other organ). I guess what I'm really wondering
is "where" in the genome we can expect to find the difference between
one purported mental module and another, and how this encoding scheme is
not hostile to the evolution of new mental modules by natural selection.
Since we don't really know, can someone explain at least what we might
expect an answer to this question to look like?
- Emil G.
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:40 MDT