Random AI geekiness about textbooks

I'm just lying on the couch with Will listening to music and reading my friends' blogs --- not something I do often! And an old posting on Geoff Knauth's reminded me that I have now met both Stuart Russell and Peter Norvig, who together have written my (and most people's) favourite AI textbook. Their first edition is one of the two textbooks I've ever read cover to cover (the other being the fourth edition of Carlson's "Physiology of Behavior"). Will and I own three copies of the current edition for reference at home, work, and one to loan grad students.

I got to talk to Peter Norvig at Google for 20 minutes (that stretched to 30ish) thanks to my former flatmate Pearl Renaker (ne'e Tsai), which was just fantastic, he is the coolest guy. I knew this from his web page, that I love the way he thinks, but people aren't always as cool in person as they are on the net. I say all this although sadly I've failed to get any research funding from Google so far. Anyway, I just met Stuart Russell at the NIPS workshop earlier this month, though I didn't really talk to him. But at least I have a sense of him now.

Of course I got to work with Patrick Winston at MIT, who also has an excellent series of AI textbooks. I got to TA his undergrad AI course, and better, attend some of his group meetings. He is also very nice and interesting, but I think I am closer to sharing worldviews with Peter Norvig. I also briefly met Geoge Luger, whose AI book they used at Edinburgh back when I went there in 1991, in Brendan McGonigle's lab.

I can't imagine writing a textbook. I just don't attend to details enough, and they are too long for me to even sit down and read and double check. It must take massive organization. Probably 10% of any textbook I ever wrote would be backwards from what I'd meant to say! I normally have to go over my good journal articles about 10 times to make them say what I intended.