This isn't news to you, but people get really exercised about ideas. Philosophical debate is almost a contact sport. In fact, you've been referred to as a "philosophical bruiser."

Philosophy is about issues that matter desperately to people. Years ago I wrote a refutation that made a lot of people mad. I refuted what I call "Strong Artificial Intelligence." This is the idea that the brain is just a digital computer and that consciousness is just a computer program running in the brain. You can refute this Strong AI thesis in about two minutes.

Ready? Go.

Here is how it goes. Let's get a watch so you can time me. I don't speak Chinese. I can't understand a word. It just looks like squiggles to me. Now, imagine that I am locked in a room and people give me questions in Chinese. I go through the steps of a computer program written in English to see what I'm supposed to do in order to answer the questions. I respond to the questions with the appropriate Chinese squiggles. I give the right answers. I do what a computer does. But I still don't understand a word of Chinese. And the conclusion is that if I don't understand Chinese solely on the basis of implementing the Chinese-understanding program, then neither does any other digital computer, solely on that basis, because the computer has nothing that I don't have.

Computation by itself isn't the same as thinking because computation is defined in terms of syntax, specifically the manipulation of symbols, which are usually thought of as 0s and 1s but could be Chinese squiggles or anything else. But real understanding is more than just syntax; it is also semantics. To understand something, you have to have meaningful semantic content.

You dispensed with the argument pretty quickly, but is there any utility for you in engaging in these sorts of debates, even if you find the arguments on the other side weak?

The beauty of arguing with people about artificial intelligence is that its adherents are committed to rationality. They may cheat like crazy in actual argument, but they recognize that there are steps to the argument, and they recognize that there is a distinction between valid and invalid arguments.

In philosophy, part of your task is to make sure that the truth has a chance of prevailing. That means you've got to try to refute obvious falsehoods. Now, it's an obvious falsehood to say there's nothing to my mind except 0s and 1s. I felt I had to refute that. Though I don't think it's the most important thing I've done, the Chinese Room Argument is probably my best-known work.

Is this leap, from syntax to semantics and from computation to meaning, the same problem as the problem of programming common sense into a computer? Why is common sense so hard to replicate artificially?

This is a different but also interesting problem. The assumption behind traditional AI is that all our thinking is done the way a digital computer operates—by a linear manipulation of symbols according to an algorithm. But there is a lot of thinking that isn't done that way. The assumption that these AI guys make is that all our reasoning is done by mathematical algorithms. It's not. A lot of thinking and acting are done using what I call "background abilities." You just have certain skills.

Here's my favorite example: My dog Russell was really good at catching tennis balls bounced off walls. Suppose you're going to build a robot dog. If you use traditional AI you would have to program the robot—this is what they think Russell did—so that it would compute the trajectory of the ball by solving a set of very complicated equations. Do you think Russell did that? I don't. He did what I would do. He consciously tried to figure out where that ball was going and then put his mouth there. He developed a skill.

page 2 | 5