When I arrived at the artificial intelligence laboratory... It was kind of the golden age of AI. Everything seemed possible. So we were just getting computers to pass the first IQ test, the first analogy tests. Or we were just getting the first computers to reason about structures of blocks and where they would fall over or doing the first language translation. Or playing, you know, they'd beat people at checkers. And we knew that they would beat people at chess.

And it was still a time when it wasn't obvious to other people that computers could be smart. So I remember there were a lot of philosophers, like a guy by the name of [Hubert] Dreyfus and another by the name of [Joseph] Weizenbaum who had arguments. I remember Dreyfus had this argument, well computers can play checkers because that can be done by simple mechanical means. But chess requires true intuition and intelligence. So a computer will never be able to play chess because it requires human intuition. And lots of people believed him. And of course we knew it was nonsense, but computers just weren't fast enough to play chess yet.

And I don't know what he ended up saying when, you know, computers became... Beat [Garri] Kasparov, but he probably shifted to, 'Oh, well yes, but emotions require human intuition.' So there's lots of things that, you know, people, when they don't have the imagination to see that the machine could do them, they imagine they have some magic human quality that allows them to do it. And so there's a kind of... It's very analogous to vitalism in life where there was a long time where people didn't believe that life could be made out of just organic molecules. They believed it had to have some special ingredient to make it alive. There's still lots of people who believe a machine can't be really intelligent, it has to have some special ingredient like consciousness or something that's some magic elixir that comes from the outside. But I think it's just nonsense. I think that special ingredient emerges from the interaction of all the interaction of simple things like life emerges from the chemicals, consciousness emerges from the computation. But at that time in AI there wasn't much evidence that that was true, but it was looking really good because everything that we were trying, like vision and speech translation and robots, it was all making huge progress very fast. And so there was kind of a sense that if we just worked on it a bit harder, everything would just work.