I think you two are basically arguing over whether the glass is half full or half empty, while the glass in question sits unseen in a locked box in an IBM research park. It is unquestionably true that humans are not performing a complete and exhaustive search down the tree of possible outcomes. The short-term memory buffer limitations of people (even in grandmasters who have greater STM chunking abilites in this domain due to practice) see to that. To the extent that Deep Thought relies on such an approach, then Deep Thought is "un-AI-ish". Now, as you point out, Deep Thought does use strategies for pruning the tree; otherwise it would be unmanageable, even for a supercomputer. Still, within that tree, it still can run through comparisons to a degree so far above that of humans (e.g., without any such thing as STM loss during the comparison process), that this "innate skill" doubtlessly contributes a lot to its ability to play well. Which doesn't sound like human players. Though the tree-pruning part does.

If Abigail's point is just the second sentence back, then he's right. If your point is the following sentence, then you're right, too. Personally, I'd be interested in seeing a detailed analysis of the importance of the strategies vs. what I've called the "skills" of D.T. towards its ability to play well, vis-a-vis human data and well-supported theories on human chess performance. Including a comparison of what you might call "memory-handicapped" versions of D.T. vs. human players. I don't know if any of this sort of thing has ever been published, but I kind of doubt it.

-- Frag.
--
"It's beat time, it's hop time, it's monk time!"