> The successes of AI that I see (eg OCR/speech recognition and Chess/Jeopardy) use these same old algorithms with only marginal improvements and more CPU.

Backpropagation training wasn't introduced until 1986 (http://www.nature.com/nature/journal/v323/n6088/pdf/323533a0...). SVMs weren't useful until the kernel trick was applied to them in 1992 (http://dl.acm.org/citation.cfm?doid=130385.130401). Feature learning wasn't an active area of research until the 2000s.

There have been huge improvements in algorithms since the 1960s. The only things around back then were a few speculative papers on analytic methods. The current state of the art in learning algorithms is a huge advance over just having some ideas about the mathematical properties of learning and a few analytic tricks in obscure papers.