7 THE RISE OF THE LEARNING MACHINES

One of the strange things about the early decades of artificial intelligence (AI) was that learning was largely ignored. Researchers were writing programs that could recognize visual scenes, perform logical reasoning, or plan a complex task, but learning was not a big part of these systems. The dominant view seemed to be that AI was a kind of puzzle, or set of puzzles, that might be cracked by finding the correct solution or solutions. And the different dimensions of intelligence could be dealt with through separate projects for each puzzle. Hence, AI was split into various streams, and a typical list of AI topics would cite automated reasoning, problem solving, knowledge representation, planning, learning, natural-language processing, visual perception, and robotics. Learning was included as a problem area, but it was usually treated as a generally desirable property of AI systems rather than as a topic on a par with the other subjects of study.