AI science vs. applications: accomplishments, failures and long-term goals

The author expresses the belief that the central problems in advancing artificial intelligence (AI) systems continue to be problems of representing knowledge, especially world and common-sense knowledge, extracting knowledge from sensory data, problems of knowledge acquisition and learning, understanding plausible reasoning, such as approximate deduction, induction and analogy, and related problems. The author suggests that AI is a very successful discipline despite the failure some not-carefully-expressed predictions. It needs steady, long-term-oriented support, although perhaps not necessarily at a very high level. The expectations need to be toned down, and intensive and imaginative research should continue unabated. Potential benefits to society are very high