by Hans Moravec

isbn: 0195116305
subject: General Science, Computing, Nonfiction
finished: 11/25/1997

I liked Moravec's 1988 book, Mind Children, so I jumped at a chance to see him talk as part of UVa's "Digital Directions" series a few weeks ago. The talk was good, and I managed to borrow a preprint of Robot: Mere Machine to Transcendent Mind (called Robot, Being at the time). Like Mind Children, it's about the future of robotics, artificial intelligence, and the universe.

The thesis of the first part of the book is that the main reason AI has so far produced less than spectacular results is that computers are still weaker, computationally, than the human brain. If Moore's law continues to apply, human-equivalent computers should be cheap a few decades into the 21st century. Of course the matter of programming them remains, but with massive computational power, brute force solutions could appear to be smart. Moravec cites Kasparov's sense of Deep Blue's intelligence during their matches last May -- Kasparov's feeling is the most compelling argument for Moravec's position that I've heard.

The second part of Robot is pure, mind-expanding extrapolation about a future that includes the ever-accelerating development of mankind's super-intelligent children and grandchildren. It's more about physics than computer science, and more about philosophy than physics -- it reminds me of Vernor Vinge's singularity and Freeman Dyson's excellent essay Time Without End: Physics and Biology in an Open Universe. Robot is somewhat technical but assumes no specific background knowledge; I recommend it. Update (12/98): the hardcover edition of Robot is now available in bookstores.