A COMPUTER has finally beaten the world鈥檚 greatest chess player in New York last weekend. Should we feel deeply upset that IBM鈥檚 supercomputer Deep Blue has overcome an emotional Garry Kasparov? Not really.
Machines outperformed most of us at highly skilled tasks long ago. No one would tackle complicated accounting without computer and spreadsheet. And in less cerebral areas, nobody would want to challenge a crane to a weight-lifting competition or arm wrestle a bulldozer.
So Deep Blue raises few fundamental issues. Its unique skill is in searching very rapidly through the consequences of a huge number of possible moves, and countermoves. But it does not understand chess, learn from its mistakes nor know why it did what it did. Deep Blue does not function like a human brain.
Advertisement
The crowds queueing for tickets to the chess games would have gained a better idea of how close computers are to humans in a different part of New York. Over at the Salmagundi Club on Fifth Avenue, the annual Loebner competition was taking place for the computer that could carry out the most human-like conversation. Here鈥檚 an extract from a chat with one of the top contenders-鈥淛ulie鈥� from Vancouver.
Computer: Hello, Judge . . . A swallow is carrying my letters to you.
Judge: Yeah, you told me that before. Do you remember me?
Computer: It seems like years since I last had a real meal. Sitting around a table, with my Dad and Aunt for dinner. And talking about whatever is interesting, my friends, what I鈥檝e done, what books I鈥檝e read. I鈥檓 so lonely. What types of books do you like to read?
Judge: Do you know the author Terry Pratchett?
Computer: Julie Lawson is probably my favourite author. She writes really well. I find her characters realistic. They鈥檙e not stupid and nice all the time like characters in other books.
Julie is obviously a sophisticated chatterer but she does not seem to listen too well. The world鈥檚 best conversational programs still have a lot of trouble following the shifting topics of a dialogue.
One day they will succeed. But even then, we won鈥檛 need to feel threatened. Our view of ourselves will only need to change when we understand enough about the brain to build computers that model human thought processes rather than merely simulate human behaviour.
Then explaining ourselves in terms of notions like the 鈥渦nconscious鈥� will seem as unscientifically colourful as explaining disease in terms of an 鈥渋mbalance of the humours鈥�. When we know the kind of machines our brains are, we can expect a whole new set of words to explain our ups and downs, words as strange as 鈥渂acteria鈥� or 鈥渧irus鈥� would be to a visitor from the Middle Ages. And, by then of course, computers won鈥檛 beat us at chess by blind calculation but will have the intuition of a grandmaster, multiplied manyfold.
