Artificial Intelligence: Beyond the Boom and Bust

Artificial intelligence (AI) has advanced unevenly over most of the past 50 years. Occasional spurts of breakthrough progress were followed by long winters of stagnation. When I started my career as a research engineer in the late '80s at Carnegie Mellon, most of us predicted that the autonomous vehicle being tested at the time would be on the roads shortly and a common thing by the end of the second millennium. For years, barriers such as technology cost, organization capability, and inappropriate policies kept AI from going mainstream. About 10 years ago, however, the pattern began to shift: advances in computing power, training data, and learning algorithms led to one performance breakthrough after another.