Kubrick was silent for quite some time, and then said,"Oh, I forgot about that.."

Stork: Did you work with Clarke too?

Minsky: He lived for a couple of weeks in the top floor of my house in Massachusetts, and we talked about a lot of science fiction -- but stories other than 2001, such as his book Against the Fall of Night, about a city run by machines. The city is great. The computer knows where all its atoms are and cleans up the city every night; I thought it was very clever.

I asked Clarke about language, and although he was optimistic about artificial intelligence, he thought that natural intonation in computers was harder. He wanted to know how long it would take computers to speak with proper, natural intonation. Gort, the superintelligent machine in The Day the Earth Stood Still, couldn't speak well.

Stork: But might that not be cinema's way of letting the viewer know this is a robot?

Minsky: Yes! It is a convention, just like the way cinematographers slow down the motion to let you know that the action is really fast! Every kid can pretend to be a superfast athlete, like a karate master, by moving really slow! I suppose we owe all this to Eadweard Muybridge who took the stop-action shots of a horse running in order to show that all four feet were off the ground at the same time.

We can credit Clarke with inventing the "fifty-year horizon": that it was really out of the question to predict what the world would look like in fifty years. He had this wonderful quote, something to the effect of "any sufficiently advanced technology is indistinguishable from magic."

Stork: Yes, and the film only described about a fifty-year jump into the future. Kubrick went to extraordinary lengths to develop remarkable special effects, all to make the film look realistic -- not the wizardry of exploding Death Stars, and so forth -- in part, I suppose, to make the audience believe in the plausibility of the basic premise, that an alien civilization had visited the earth and moon.

Minsky: Yes, and whereas some films, and Star Trek in particular, insert technobabble, Kubrick and Clarke were very careful to stick as closely as they could to science.

I was once on the set of Star Trek -- the TV version, not the film -- and spoke to Gene Roddenberry, its creator. I said that with the tremendous, large, and loyal following the show had, wouldn't it be a great opportunity to insert even just a little "real" science?

Gene thought for a moment and then said, "no -- too dangerous."

Stork: It is clear that some of the problems have been solved, chess for instance. However, we can presume that there is no Deep Blue on board Discovery -- that's too much hardware just for the crew's amusement. But we have no program that has general intelligence that can play chess reasonably.

Minsky: But we could have! Only a small community has concentrated on general intelligence. No one has tried to make a thinking machine and then teach it chess -- or the very sophisticated oriental board game Go. So Clarke made the same mistake that Herbert Simon made in the sixties, that AI was progressing so well and would continue to progress that we should start by concentrating on particular problems, such as Go or chess. But that's the wrong idea. The bottom line is that we really haven't progressed too far toward a truly intelligent machine. We have collections of dumb specialists in small domains; the true majesty of general intelligence still awaits our attack.

To give you one idea of some of the dumb things that have diverted the field, you need only consider situated action theory. This is an incredibly stupid idea that somehow swept the world of AI. The basic notion in situated action theory is that you don't use internal representations of knowledge. (But I feel they're crucial for any intelligent system.) Each moment the system -- typically a robot -- acts as if it had just been born and bases its decisions almost solely on information taken from the world -- not on structure or knowledge inside. For instance, if you ask a simple robot to find a soda can, it gets all its information from the world via optical scanners and moves around. But it can get stuck cycling around the can, if things go wrong. It is true that 95 percent or even 98 percent of the time it can make progress, but for other times you're stuck. Situated action theory never gives the system any clues such as how to switch to a new representation.

When the first such systems failed, the AI community missed the important implication: that situated action theory was doomed to fail. I and my colleagues realized this in 1970, but most of the rest of the community was too preoccupied with building real robots to realize this point.