<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD><TITLE>Message</TITLE>
<META http-equiv=Content-Type content="text/html; charset=us-ascii">
<META content="MSHTML 6.00.2900.2180" name=GENERATOR></HEAD>
<BODY>
<DIV><SPAN class=314175615-09122005><FONT face=Arial size=2>Hi
Neil!</FONT></SPAN></DIV>
<DIV><SPAN class=314175615-09122005><FONT face=Arial
size=2></FONT></SPAN> </DIV>
<DIV><SPAN class=314175615-09122005><FONT face=Arial size=2>The field has very
likely moved beyond my research topic by this point (I haven't been following it
for 10 years or so). The concept was to experiment with evolvable
languages for expressing structure (for example, the structure of neural
networks or other parallel processing networks with semi-uniform
architectures), in particular where the structures unfold from a "developmental"
process. The hope was that atomic operations on representations of
developmental processes would have the ability to cause both coarse and fine
alterations in structure and thus function and with an appropriate
representation would thus be fruitful for either automatic "genetic algorithm"
style exporation of the space of possible structures, or for direct "intelligent
design" by a person.</FONT></SPAN></DIV>
<DIV><SPAN class=314175615-09122005><FONT face=Arial
size=2></FONT></SPAN> </DIV>
<DIV><SPAN class=314175615-09122005><FONT face=Arial size=2>I spent a lot of
time working on formal concepts of evolvability that didn't produce any
breakthroughs; I spent a lot of time working on potential "substrates" (the
network components being assembled by developmental processes), that didn't lead
to anything particularly revolutionary; I spent a lot of time playing with
simple ideas about these languages; that led to some toy-level results that
would have probably been enough for me to continue with the research, but I was
discouraged because to me the research was basically a failure, so I gave up and
went to work in non-AI industry.</FONT></SPAN></DIV>
<DIV><SPAN class=314175615-09122005><FONT face=Arial
size=2></FONT></SPAN> </DIV>
<DIV><SPAN class=314175615-09122005><FONT face=Arial size=2>I don't have any
particular thoughts on future directions, but I think the time is getting riper
and riper for some actual progress to be made; along with increased processing
power I expect to see interesting work come to light, which is why I'm starting
to get interested in following the field and maybe dabbling a bit as an outsider
hobbyist. I think that within 5 years or so it should be possible for any
AI researcher to have access to mouse-level processing power (roughly a
milli-human), which might be enough to produce some interesting results. I
think it's really hard to do much useful work on AI with sticks and rocks and
Cray XMPs and other inadequate tools.</FONT></SPAN></DIV>
<DIV><SPAN class=314175615-09122005><FONT face=Arial
size=2></FONT></SPAN> </DIV>
<DIV><SPAN class=314175615-09122005><FONT face=Arial size=2>Good luck with your
research! What are you working on?</FONT></SPAN></DIV>
<DIV><SPAN class=314175615-09122005><FONT face=Arial color=#0000ff
size=2></FONT></SPAN> </DIV>
<DIV><FONT face=Arial color=#0000ff size=2></FONT> </DIV></BODY></HTML>