Artifical Intelligence and the Turing Test

While reading a blog entry about Artificial Intelligence this morning it suddenly struck me how the Turing Test, a commonly advocated test for determining whether a computer program has achieved intelligence, is like supreme court judge Potter Stewart’s test for pornography, “I don’t know how to define it, but I know it when I see it.”

In the Turing Test a computer program converses by teletype with a human being who does not know whether it is a human or computer on the other end of the line. The computer program passes the test if the humans with which it interacts thinks they are talking to another human being.

It is strange and telling that no better test for intelligence has been proposed, since this is a test for imitating human beings rather than intelligence itself. On the original Star Trek television show an alien named Dr. Spock was portrayed as much more intelligent than humans in many ways but his lack of human type emotions and experiences was a constant barrier between him and the rest of the crew. It is obvious that as intelligent as he was he would never pass the Turing Test. As much as he desired to have a good relationship with the other crew members his inability to simulate human emotions enough to pass was a constant irritant.

As a computer programmer I know that completely understanding the task to be programmed is necessary to successfully create the program. The less completely you understand the problem and its possible solutions the less successful your program will be. It is true that in some situations you can start off without complete knowledge and learn more about the problem as a result of trying things that fail. Eventually you may gain enough knowledge through repeated failures to finally succeed but at that point you understand the problem and solution thoroughly.

When we fully understand intelligence and consciousness we will be able to create computer programs that are intelligent and conscious and not one minute before. When we reach that point I am sure that we will have much better tests for intelligence than the Turing Test. There is much more to intelligence than acting like a human being.

Don’t talk to me about Artificial Intelligence

My wife was at a function the other night without me and ran into an acquaintance. ‘Your husband is a computer programmer isn’t he?” she asked my wife. “I’d like to talk to him. I just read something about Artificial Intelligence and would like to talk to someone about it.”

“You don’t want to talk to Dave about Artificial Intelligence,” my wife replied.

The reason no one wants to hear my views on the subject is that I don’t believe in it. Among people who are interested in and talk about Artificial Intelligence some are very happy and optimistic about the idea that computers some day may became so intelligent that they no longer need human programmers to tell them how to solve problems. Others say they are fearful and apprehensive about that possibility. It turns out neither group wants to be told that it won’t happen.

You might think that the people who claim to be very worried about something would be relieved and grateful to be reassured that it won’t happen, but when it comes to worries about a future in which computers are more intelligent than their human creators you would be wrong. I’ve given the matter some thought and decided that there are two possible explanations. The first is that people like the feeling of being worried and apprehensive, at least about an exotic and intriguing future. The other possibility is that they are not really fearful and saying that they are is just some sort of conversational gambit – a role in some sort of unscripted human drama in which all the world is a stage and everyone is a player.

I won’t tell you the reasons why I don’t think Artificial Intelligence will ever exist, at least with the kind of digital computers we have now, because, like I said, my wife tells me no one wants to hear my views on that subject.