Artificial Intelligence

Part I

Artificial Intelligence. We’ve been surrounded by the concept for a while now. Movies and television shows have a lot of ideas about it, and sometimes it’s neat. And sometimes it’s dangerous. From robotic cops to homicidal houses; from artificial children to perfect wives.

I’m not a big fan of the term “artificial intelligence.” Either something is intelligent, or it isn’t. I’m not going to nitpick, though, because there’s nothing else it can really be called. Faux Intelligence, quasi-intelligence, whatever. It’s all the same.

These days, AI consists of a system of canned responses. You type something to the computer, it looks through a database of words, and comes up with something that might be relevant. Like the old program Eliza. Certainly, they’ve gotten a bit better since then (it would be hard not to) but they’re still not up to snuff.

Personally, I don’t believe it’s possible to achieve the Hollywood concept of AI. Not right now. I do think it would be possible to come up with something that would pass the Turing Test. That would be relatively easy. And I do think it would be possible to achieve something very close to true intelligence in a man-made object. Just not right now. This, unfortunately, brings to mind some serious questions about us, humans, and our intelligence.

This is a general description of the Turing Test:

“When talking about the Turing Test today what is generally understood is the following: The interrogator is connected to one person and one machine via a terminal, therefore can’t see her counterparts. Her task is to find out which of the two candidates is the machine, and which is the human only by asking them questions. If the machine can ‘fool’ the interrogator, it is intelligent.”

When worded this way the error of the Turing Test is glaringly obvious: just because the computer fooled somebody does not mean that it is intelligent. How well the computer fools someone is entirely dependent on how thorough the program’s designers were. Asking an innocent question that the designer didn’t think of could blow the entire deal. “Say, where did you grow up?” or “What high school did you go to?” Still, that would just blow the test. You know the computer is still not intelligent. It’s not thinking. If you asked it “How are you today?” it will not try to figure out how it is, judge whether you’re really interested or just making conversation, and then come back with a suitable reply. It will compare your question to a set of answers it’s been programmed with and then respond back. It may fool somebody, but it’s not really thinking about it.

But then, that’s why it’s artificial intelligence. It’s not real. It’s fake. The Turing Test, then, boils down to a type of IQ test for the designers rather than the computer. And the people asking the questions.

As an aside, I’ll mention here a little story about something that happened to me when I was in high school. We had a class trip to the AT&T building in New York City. They had a big display there about technology and computers and things. Kind of like an amusement park for nerds. One of the things they had were two terminals that were hooked together, back to back. Two people could talk back and forth. Now, with instant messengers it’s no big deal but back then it was neat. Anyway, I was typing to some girl and she thought she was talking to a computer. To convince her otherwise her friends had to move her around to my side and see that I was, in fact, typing what she was seeing. I suppose it doesn’t say much for my personality but I am proud that I managed to pass the Turing Test.

Advertisements