Woody Bledsoe

"Incidentally my goal wasnt to necessarily press forth AI. It was out of these asides, but AI was used to prove theorems early on. Jerome Simon had one of the first provers that I considered a good AI. Imply was supposed to do the same kind of thing. It was suppose to be an interactive prover and a stand-alone prover at the same time."

"Partly, but partly to see how far we could go with automating the proof of mathematical theorems. That was the intent."

"No, other AI people who werent working on theorem provers in particular. I dont know if I can give a lot of references off the top of my head, but that was the state of the art for awhile. It used heuristics and that turned out to be later not as popular amongst the formal automated theorem proving crowd to use heuristics. They wanted to find out what you could do to stand alone heuristics."

"We proved — in the stand-alone mode with heuristics — we proved limit theorems of calculus and thats a good sample"

"Heuristics are ad hoc. They tend [correct word?], you fix up the heuristics for limit theorems. Now when you go to algebra you need some new heuristics. And when you go to functional analysis and you need some new ones. When are you going to stop? This is what we want as a general purpose prover here. It will prove anything. Youre in a dilemma. I mean if you get a general purpose prover oftentimes theyre too weak."

"You get a very special purpose prover that uses heuristics, then they stop at the boundary where they cant work anymore."

"I dont think its too hard in any particular area, although I think people are a little surprised how easy it was in limit theorems. I dont think the average person would do it."

"It was built to partly prove theorems, which is one of the jobs that AI does. [inaudible word] emission, theorem proving, and so forth."