Modes of AI research: stealing old ideas

“This time is different” is a common refrain from AI’s more passionate advocates. Here’s Maggie Boden on the history of ideas in AI, from her incredible 1,000+ page history of cognitive science.

If today’s AI researchers aren’t deliberately stealing and renaming old ideas, as Father Hacker advises, they are sometimes reinventing the wheel—and lack of historical knowledge doesn’t help. For instance, 1980s–1990s work in machine learning often replayed insights available in traditional statistics (see 12.vi.f, and iii.f above); and the NewFAI researchers Michie and Gregory both said that the best model of the real world is the real world itself, long before the situated roboticists did (iii.b, above).

Boden 2006 Mind as Machine: A history of cognitive science volume I and II OUP. p 1102

As part of the SCALINGS project, we’re looking at how technologies like self-driving shuttles are being revived for today’s testing environments, with often little or no historical knowledge amongst engineers of their past successes and failures, or the social and technological consequences of those failures.