We will begin with the first chatbot, the famous ELIZA, which "is based on a "script" consisting of patterns and corresponding responses." This explains why it responds the way it does, sometimes seemingly unrelated to the previous statement; if it recognizes a keyword, it will search through its list of appropriate responses and pick one. One quip about that is if you enter the same statement over and over, you will eventually begin seeing the same pattern of responses from it, its stores exhausted. I thought it was rather amusing. Of course, it is also fashioned after a Rogerian psychologist, which may also add to the repetitious replies of support - it makes the copying of my words almost acceptable. I can agree that ELIZA does that well. But for the same reason, after a few minutes of chatting, I was bored and you could say annoyed with ELIZA; nothing personal, but I just grew tired of being asked question after question and for more elaborations. So then I moved on to the next chatbot.
ALICE was different from the get-go, with its own little avatar on the left!! In addition to that, ALICE was also different from ELIZA because it asked questions to learn about me, but not in the same way that ELIZA did. ELIZA was trying to figure me out, more or less, like a psychologist, where ALICE was simply making conversation. ALICE would ask some weird questions about "Om" though, and it sometimes shouted things like "DIALOG HISTORY". ALICE felt more human because it would try to talk about things that it wanted to, instead of the one-way conversation with ELIZA. What made it less human was when it would blatantly say it was an artificial intelligence, and still the way it talked. I could see how the 'intelligence' was growing from each chatbot. After talking with ALICE for a bit, I almost missed ELIZA.
Finally, there was Jabberwacky, which was the wackiest of them all. Certainly, Jabberwacky seemed the most human, not only with its inclusion of spelling mistakes and casual way of talking, but also with the tone it expressed, an ounce of what could be feeling poured into each of its responses. But could this be actual intelligence? Or was the tone simply implicit in the reply it responded with? I believe that the set of responses for Jabberwacky were just programmed much better than those of ALICE and ELIZA. Perhaps the way it could select its responses was also more advanced. Of the three, Jabberwacky was the 'most realistic', and it may have passed the Turing Test for me if I hadn't already know that it wasn't a person.
However, from what I've seen, I don't think these artificial intelligences are truly thinking yet. Certainly, they need to select which of their responses to spit out at the users, but isn't that more like an algorithm? Thinking involves much more, in my book; it has to do with learning, but more importantly, understanding. As with the Chinese Room, because the person can relay perfect Chinese doesn't mean that he (or she) can necessarily hold a conversation out of the room. He hasn't learned the language, he merely copies something and writes it down; he doesn't know how to speak it.
When a program has a specific goal in mind, and can learn, can adapt to different obstacles in its path and attain the goal, that's when I think programs will be thinking. I think the chess-playing computers are closer to thinking than the chatbots, or at least the three we were given (I've heard about a chatbot called Cleverbot that is supposed to be pretty..clever). True, the chess computer has a list of moves that it can execute, but as we learned in the class, the list for a whole game, with every move put into account, is much too huge for it to run efficiently. And so it looks ahead, we could say, five turns. From there it keeps modifying its list of moves with respect to each one the human player makes until it wins or the human wins. I say this is much closer to thinking because it requires a sort of learning; no two games of chess will end up being the same, but there are trends that the computer can recognize and therefore remember for future reference. Just like 20 Questions. That step, where we are now, is the one right before what we are striving for; a fully independent , fully functional, fully thinking program.
Talking about this made me think of the Pixar film, The Incredibles - specifically, the Omnidroid robot the family fights at the end, constantly learning and adapting.
Until we can build an A.I. that can do that, I don't believe what we're making is thinking. But then you have to wonder, would we have a robot revolution on our hands?
