Blogue-Blague

Minsky & Dennet Meet on the Brain

From Wired a great but short interview:

WIRED: What's wrong with the traditional approach to how the brain works?

Minsky: Physics gives us about five laws that explain almost everything. So we keep looking for those kinds of simple laws to apply to the brain. The idea in my new book is that you shouldn't be looking for a single explanation of how thinking works. Evolution has found hundreds of ways to do things, and when one of them fails, your mind switches to another. That's resourcefulness.

In The Emotion Machine, you argue that feelings result from switching on or off certain "mental resources."

Minsky: The traditional view of emotions is that they are something extra, like adding color to a black-and-white photograph. But to me, emotions are what happens when you remove other resources. Anger means you've turned off your social graces, you've turned off your cautiousness, you've turned off your long-range plans and most of your ambition, and you've turned on things that make you act more rapidly and less deeply. Recognizing this complexity adds dignity to the theory.

Dennett: Computer programmers have the luxury to create hierarchies of control. The systems, the subsystems, the sub-sub-subsystems are complete slaves. They never rebel. This gives you a model of the mind with the highest echelons of logic at the top. But if you think about a brain as a community of individually semiautonomous, even independently evolved agencies, as Marvin has, you realize that the agencies have to be browbeaten and they have to form alliances. Emotions aren't an add-on but rather the politics of the whole system.

So what would a machine that worked this way look like?

Dennett: Like us.

Minsky: A well-designed program that wouldn't be so hierarchical but more like a network with resources that make requests of other resources.

Dennett: The research world is going to be impatient with Marvin because they are eager for computational models that really work. Marvin is saying, "Wait a minute, let's work out some of the high-level architectural details in a way that's still very loose, very impressionistic. It's too early to build the big model."

Minsky: Actually, I could quarrel with that. I think the architecture described in The Emotion Machine is programmable. If I could afford to get three or four first-rate systems programmers, we could do it. You can get millions of dollars to drive a car through a desert, but you can't get money to try to do something that's more human.

Why is the idea of a thinking machine so compelling?

Minsky: I think there is a worldwide survival problem. As the population grows and people live longer, there won't be anybody to do the work. So there is an urgent need to make inexpensive mechanical people that are able to do all the things that moderately unskilled people do now.

Dennett: I don't find that very convincing, Marvin. I think we're interested in it for purely curious, scientific reasons. We want to know how we work.

Minsky: Or make machines that work better than us and can solve all the problems we wanted to but couldn't. As Hans Moravec used to say, "The machines will be us."

Dennett: Marvin, I have a slogan for you that came to mind while I was reading your book. I've used it myself as a paper title. "My body has a mind of its own, so what does it need me for?"

Minsky: [Laughs.] I once peeled a label off a London bus. It read: MIND YOUR HEAD.