Humans have just been beaten in our own game. Valentine's Day 2011 will go down in history, not because of the record amount of greetings sent via social media, but because of the first airing of Watson taking on two guys in Jeopardy. Like chess master Garry Kasparov, the two human contestants are famous champions in their game. Kasparov was awed in his first match against IBM's Deep Blue in 1997, because he "could smell a new kind of intelligence across the table". In fact Deep Blue merely calculated chess positions really fast and was relatively stupid, but the new Jeopardy champion Watson is something more remarkable. It's amazing how human-like the answers from a computer can seem. This is Wolfram Alpha taken to the next level.
Automated general knowledge processing has some really far-reaching implications. If I ever fell seriously ill, I would much rather go to a doctor with a machine like this compared to a doctor without. If I ever had a large amount of money to invest, I'd rather take advice from IBM than from anyone on Wall Street. What does that make of IBM as an investment, I wonder? Daily arbitrage-seeking dealings in the financial world are one area where specialized AIs have already largely taken over. No living person could ever hope to learn as much as these systems are able to learn in just a year or two, and no living person could ever hope to make informed decisions as fast. Watson's grasp of language might not be too complex, but the width and speed make up for that.
IBM has really scored a victory with this one, not just for themselves but for the whole industry, and made true artificial intelligence look that much more feasible. It's been 14 years since machines took over humans in chess. Will another 14 years be enough for IBM to bring an even more human-like AI to the table, maybe competing on American Idol instead of Jeopardy? In my last post I talked about the hopes and possible slowdowns in the growth of computing power, and those can really affect this kind of development. But what about after 25 years, if new technologies have broken through by then? Maybe a machine could pass the Turing test by 2035, being indistinguishable from a human via a text-based medium. Maybe algorithmic development will be slower than the development of computing, and we will only see machines finally beating humans in Go. But even if they fall short, the likes of Watson will affect a major part of our lives regardless.