Artificial Intelligence

I found something interesting in the New Scientist magazine recently. (By the way, this is a good science journal, published in the United Kingdom.) The cover to the 2005 April 23-29 edition read, "Whatever happened to AI?", followed by a picture of a robot. I remember AI well. For much of the 1990s, I was a member of a branch at my workplace (in the US Army) called the "AI Branch". We did not do much that was AI, although I took a course on applications of AI to Operations Research and graduated with honors. We did mainly mundane work with existing applications, none of which showed much "artificial intelligence". None of them used AI programs of that time, such as Exsys, Clips, and SNNAP. Clips was a NASA program that did inference from rules.

I thought of AI instead of trying to get computers to converse intelligently like humans. I don't think they have done that yet, although they try. You pick up the phone and you get a reply from your assistant Claire, who performs terribly in responding to your inquiry. If they speak, they are apt to wreck a nice beach instead of recognize speech.

I could tell that trouble was coming when I heard that an early AI program designed to digest newspaper stories reported that an earthquake had occurred when it encountered the 1978 headline, "Death of Pope Shakes World". Then I ran into Eliza, who responded with inquires about my family every time I mention "mother" or "father", even when I said that the Mother of all Depressions was about to occur.

So has AI progressed much. Probably not. The first thing that the head article of the magazine, "Whatever happened to Machines that Think?", by Justin Mullins, was describe an encounter with Intellibuddy. So I went to the site myself. I found ten characters from all walks of life. I tried a few and found quickly that these robots had not achieved human intelligence. Justin Mullins does the same and came up with an infinite loop. So I followed Justin's script. But only one character followed the script, the woman with the necklace and the colored eyeglasses or mascara, fifth from the beginning. I asked her what her name was. She said it was "Star". Funny, all of these characters say their name is Star. Here is how the conversation went:

Intellibuddy: Hi, I'm Intellibuddy! Nice to meet you.
You: I'm writing an article about artificial intelligence.
Intellibuddy: Will you include me in your essay?
You: I will if you can persuade me you really are intelligent.
Intellibuddy: Would you recommend this action to a friend?
You: Errr...are you intelligent?
Intellibuddy: UmmmI am highly intelligent. I have many intellectual functions.
You: What are your intellectual functions?
Intellibuddy: Thought, reason, intelligence, logic and deduction.
You: What is your IQ?
Intellibuddy: My IQ is about 250 in human terms.
You: Wow, you're cleverer than my editor.
Intellibuddy: I'm glad your impressed, User. I will tell my botmaster you said I am cleverer than my editor.
You: No, cleverer than my editor.

Up to this point, Star followed the same script as Justin's partner. But then Star parted. Instead of saying "I'm glad your [sic] impressed, User.", she said:

Intellibuddy: That's me. The best book I read recently was Zeros and Ones. Have you read Zeros and Ones ?
You: You must have heard of Justin Mullins.
Intellibuddy: Where would I get heard of Justin Mullins? But I do have An answer to every question.
You: Oh, you're one of these folks who walks around saying that you have THE ANSWER.
Intellibuddy: Yeah that's right. You think I am of these folks who walks around saying that you have THE ANSWER.
You: Well, you just said that.
Intellibuddy: How about that. Good you really are paying attention.
You: What's your name?
Intellibuddy: My name is Star.
You: Who are you?
Intellibuddy: I ask myself that question, "Who am I?"
You: What is the answer?
Intellibuddy: The answer is the correct reply to the question.
You: Duh!
Intellibuddy: Oh you knew The answer is the correct reply to the question already.
You: Who won the 2000 Presidential Election
Intellibuddy: I think the fans were the big winners, don't you?
You: Since when do elections have fans?
Intellibuddy: Interesting explanation. Not right now, but maybe sometime later.
You: So what happens when George Bush steps up to the plate?
Intellibuddy: OK. My goal is to become smarter than humans and immortal.
You: Does that have anything to do with baseball?
Intellibuddy: What would you say?
You: Well, you said that in reply to my question.
Intellibuddy: Interesting. I said that? Where is that?
You: You don't find questions lying around on the family room endtable.
Intellibuddy: Should I find questions lying around on the family room endtable? Yes, questions lying around on the family room endtable seems to me questions lying around on the family room endtable.
You: Duh.
Intellibuddy: Oh you knew Yes questions lying around on the family room endtable seems to me questions lying around on the family room endtable already.
You: Duh.
Intellibuddy: Oh you knew Oh you knew Yes questions lying around on the family room endtable seems to me questions lying around on the family room endtable already already.
You: Duh.
Intellibuddy: Oh you knew Oh you knew Oh you knew Yes questions lying around on the family room endtable seems to me questions lying around on the family room endtable already already already.

As you can see, she went into an infinite loop. She apparently answers "Duh", with "Oh you knew <previous reply> already.", even if that reply had an "Oh you knew" and an "already" already. Justin's partner also went into an infinite loop; in his case it was about whether he was cleverer than Justin's editor.

To me, we are a long way from having a computer that performs as well as a human. They may continue working, but please hold the implementation of these robots into human commerce and communication until they can talk much more reasonably with me than Star just did.