Recognizing and generalizing nontrivial mathematical patterns from the world, at the level you mean, seems (to me) to be AI-complete; on the other hand, in a domain with lots of variables and unreliable human intuition, Eurisko did pretty well at basically that.
The controversy around AM is the same as around Eurisko, right? that it’s not clear how much was the AI and how much was Lenat. I believe that Lenat says the Traveller fleet design was 60% Lenat and 40% Eurisko. It’s not clear what it means to quantify that, so it’s not clear that he disagrees with the skeptics. In fact, his abandonment of Eurisko suggests that he agrees. Here’s the OB discussion of Lenat.
Recognizing and generalizing nontrivial mathematical patterns from the world, at the level you mean, seems (to me) to be AI-complete; on the other hand, in a domain with lots of variables and unreliable human intuition, Eurisko did pretty well at basically that.
Funny that you should mention Eurisko and not Automated Mathematician by the same guy, because the postulated agent sound pretty close to AM.
Huh, I hadn’t heard of that. The controversy around it is… interesting.
The controversy around AM is the same as around Eurisko, right? that it’s not clear how much was the AI and how much was Lenat. I believe that Lenat says the Traveller fleet design was 60% Lenat and 40% Eurisko. It’s not clear what it means to quantify that, so it’s not clear that he disagrees with the skeptics. In fact, his abandonment of Eurisko suggests that he agrees. Here’s the OB discussion of Lenat.