zlacker

> In the dawn of time Lenat actually did stuff.

Harsh, but probably fair. Or as Eliezer Yudkowsky put it "EURISKO may still be the most sophisticated self-improving AI ever built [...] by Douglas Lenat before he started wasting his life on Cyc."

> I've been playing around with a port of AM for the past coupla weeks.

From where is the code for AM available? I ask because some people on an AI list I subscribe to think the Eurisko code is unavailable (but would like to get hold of it). You wouldn't have the Eurisko code too, by any chance?

> I think the world's coolest AI programs are Eurisko, Copycat and AM.

I would agree, in terms of modelling general cognition.

>>cabala+(OP)
No, eurisko nobody has :) The version of AM I have was written by one of my professors at the University of Texas Austin, and is in an ancient dialect of prolog. I'm not sure if it was ok to redistribute. Let me check.

I got it working on gprolog last week, but it seems to cause stack overflows :(

>>akkart+A
> No, eurisko nobody has :)

So why is that? Does Lenat not want anyone to see it? Or was the code lost?

BTW am I right in presuming that the original AM was written in Lisp?

>>cabala+h3
Yes, the original papers indicate it was in lisp. "Why AM appears to work" suggested that a lot of the reason math was a good domain for discovery was just that lisp is well-suited to the domain. And Eurisko could modify heuristics expressed as s-expressions.

People have asked him, but I'm not aware of Lenat ever commenting on the source code. Cycorp is substantially funded by DARPA, and I wouldn't be surprised if Eurisko is part of its IP now.

Feel free to email me to take this offline.