It's a long time ago now (late 80s early 90s), but reasons include:

- I hated the politics/game-playing that seem to be a huge component of getting on in academic research

- It was clear to me that we are unlikely to see any fundamental breakthroughs in general AI during my career

- I had been doing work on the Web since '92 and it was clear to me by '95 that this had a much brighter future than "classical" AI

- I co-founded a Web/Java startup in '95