From http://flickr.com/photos/treacletart/

Last week I mentioned that computers had come very close to pass the Turing Test. Before I explain why this year’s Loebner Prize contest was so significant and why next year’s is likely to be even more so, let me give some history on Turing and his test. In the field of Computer Science, Alan Turing is known by several different names, including the “father of modern computing” and the “father of artificial intelligence”. That alone should be enough for most people to understand that he’s a pretty important guy. What a lot of people don’t know about him, is that he broke the Germans’ Enigma machines during World War II, allowing the Allies to listen in on German communications. This breakthrough in using machines to decipher codes played a very important role in the Allies’ victory. Another thing that most people don’t know is that Alan Turing was gay. He was a gay man living in England during a time when being gay was illegal. Despite being a hero who had just saved his country, the government forced him to take estrogen as a way to combat his homosexuality, a standard treatment at the time that is now rather mystifying. In 1954, Turing killed himself by eating an apple soaked in cyanide (as you might guess, Snow White was one of his favorite movies.) So a brilliant man’s life was cut tragically short at a mere 41 years. During his life, though, he gave much to the fields of cryptology, computing, and artificial intelligence. I want to focus on that last one today.

The Turing Test is a very simple game that is played by two people and a computer. One person is the judge and they sit in front of a computer. The computer and the other person sit at computers in other rooms and the only way any of them can communicate is through text. Essentially, the three of them are in a chat room. The judge’s job is to determine which of the two others in the chat room is the computer and which is the human. The computer is trying to fool the judge, the other human is trying to help the judge, and no topic is off limits. Of course the first thing that everyone suggests is that the human simply says, “I’m the real human!” but what they forget is that the computer can do the same thing or even say, “No, they’re lying! I’m the real human.” Turing predicted that by the year 2000 a computer would be able to fool the judge 30% of the time in five-minute games. Since this test was first suggested in the 1950s, researchers have been attempting to pass this test and thus have a computer program that is considered to be artificially intelligent by at least this one, long-standing benchmark.

To help you wrap your head around exactly what I’m talking about, think back to 5 or 6 years ago when bots were really popular on AIM. If you never experienced one of those, you can also go talk to the 2004 Loebner Prize winner, Alice. Hopefully that will give you a sense for the kind of interaction between the judge and the computer. The only piece that website is missing is the second human who would also be talking the whole time.

In 1990, the Loebner Prize was created. It’s an annual contest that awards a prize to the program that comes the closest to beating Turing’s prediction every year and will award an even bigger prize when a program actually crosses the threshold. This year many programs did very well, but one managed to fool a panel of five judges 25% of the time over a series of 5-minute games. So next year we’re very likely to see a program actually pass the Turing Test. This would be a big breakthrough in artificial intelligence and would break a both an important hurdle and a long-standing psychological barrier. Once the Turing Test has been passed, many researchers will begin to think of artificial intelligence as being possible which may well lead to somewhat of a renaissance in the field.

It’s also important because it begins to raise questions about what happens if artificial intelligences are created. Are they people? Do they have human and civil rights? Can they commit crimes? Or are they owned by people and companies? If they’re owned, what will public sentiment be? Will people think of it like slavery or like owning any other piece of software? These are questions that our society may have to confront in our lifetimes and not many people outside of academia have given any real thought to them. On the other hand, the programs that are coming close to winning this prize right now are relatively simple, many of them written by a single person in their spare time. That could either mean that the Turing Test simply isn’t a good way of judging intelligence or it could mean that humans are more simple than we’d like to believe. In any case, the next several years should be very interesting for computers and AI.

Related posts:

1. Obama takes South Carolina. . . Is anyone else getting tired of this?
2. What I’m Looking Forward to in Viet Nam
3. Step Away From Religious Politics: Slowly.