Artificial intelligence began with an ambitious research agenda: To endow machines with some of the traits we value most highly in ourselves—the faculty of reason, skill in solving problems, creativity, the capacity to learn from experience. Early results were promising. Computers were programmed to play checkers and chess, to prove theorems in geometry, to solve analogy puzzles from IQ tests, to recognize letters of the alphabet. Marvin Minsky, one of the pioneers, declared in 1961: “We are on the threshold of an era that will be strongly influenced, and quite possibly dominated, by intelligent problem-solving machines.”
Fifty years later, problem-solving machines are a familiar presence in daily life. Computer programs suggest the best route through cross-town traffic, recommend movies you might like to see, recognize faces in photographs, transcribe your voicemail messages and translate documents from one language to another. As for checkers and chess, computers are not merely good players; they are unbeatable. Even on the television quiz show Jeopardy, the best human contestants were trounced by a computer.
In spite of these achievements, the status of artificial intelligence remains unsettled. We have many clever gadgets, but it’s not at all clear they add up to a “thinking machine.” Their methods and inner mechanisms seem nothing like human mental processes.
Perhaps we should not be bragging about how smart our machines have become; rather, we should marvel at how much those machines accomplish without any genuine intelligence.