When my family first got the internet I was in 6th grade. I signed up for an AIM account and was astonished by the speed with which I could communicate to all my junior high pals. Sure, we still had dial up but it was amazing. A few months after I begin talking online to my friends, I was an old pro by this time, I remember someone sending me the name of a computer that you could talk to via AIM. After the initial fun of saying cuss words (oh no!) to it wore off I began to realize that the computer was merely repeating to me a series of predetermined responses, regardless of what I said to it. I never messed with it again and pretty much forgot about it until this assignment.
That being said, when we were instructed to converse with Chatbots I wasn't expecting to understand it anymore than I had in the 6th grade, let alone be impressed by it. My experience this morning however has led me to a completely different conclusion. I will examine each Chatter in turn starting with Eliza...
ELIZA
I started talking with Eliza this morning. She first asked, Please state your problem. I said, Eliza I don't have a problem. She came back with, Are you sure you don't have a problem? Yes, I said. You are sure, she asked. Yes, came my answer. I see, said Alice. As I continued talking to her I was noticing that her responses were usually at least 50% comprised of my initial statements. Me asking, What is the weather like today? had her returning with, Why do you need to know what the weather is like today? I eventually said, Are you aware you are not human? Alice responded, We're not talking about me here - what about you?. Then it hit me. I had forgotten te little teeny tiny sentence that totally put Eliza in context: She is modeled after a Rogerian psychiatrist. Da Ding! Sure, Eliza didn't really seem to want to answer any of my questions. She was trying to get me to evaluate WHY I was asking certain questions and just how sure I was with my answers. In this way I very much realize how she succeeds to pass the Turing Test AS a Rogerian psychiatrist.
ALICE
At this point I have not talked to Alice. The website alice.pandorabots.com is down. (11/13/2008)
JABBERWACKY
And Wacky he is... What I first found interesting about Jabberwacky is that within the first few lines of conversation Jabberwacky asked ME what the weather was like today. Only because I had just tried to get Eliza to talk about this did I find it surprising. Jabberwacky also had a sense of humor. At one point I stopped talking to it because I had to answer my phone. After a brief five minute conversation I returned to Jabberwacky and said Sorry, I had to answer my phone. Jabberwacky replied Sorry, did you say something? Very funny, I responded. He then asked, Did you get my little joke there? I tried to use the emotion applicationg to change the tone of the conversation but thought it was most impressive when Jabberwacky was able to decipher my tone without it. The main difference I noticed between Jabberwacky and Eliza was that Jabberwacky would ask ME questions and it was about topics that we weren't even discussing. For example at one point Jabberwacky asked me how I felt about global warming. I told it that I wasn't sure that I believed it was really happening. Jabberwacky said, It is. You can't live without it. Then my brain kicked back to my Atmospheric Science class and I remembered that global warming IS what happens every minute of every day and keeps the earth warm enough to hold life, and keeps the right about of gasses in the air for us to breathe. The rapid increase of global warming is what I was thinking of, what the media is always talking about, and what I thought Jabberwacky was asking me about. I responded to Jabberwacky saying "Oh, I thought you were talking about the whole destruction of the planet sort of thing" and Jabberwacky said "We're not talking about Lord of the Rings here". I think that sums it up....
TURING TEST(S)
For whatever reason the Take the Turing Test portion of the assignment is down too. I believe it is also connected to the pandorabots cite that ALICE relies on. So I have not completed that yet either. (11/13/2008)
REFLECTIONS
I must admit that even after reading all of the assigned material I don't feel qualified to certainly predict the future of the Turing Test one way or the other. Will a machine be able to pass the Turing Test in ten years? Maybe. Maybe not. What I have gathered from this reading is that, in the case of the Loebner prize, collective efforts to pass a Turing Test aren't adding to the bank of knowledge in the field of AI but are rewarding bigger and better ways to effectively "cheat" the system. I don't really believe that outsmarting the Turing Test is really passing the Turing Test. I agree with the author when he discusses the pitfalls of a contest that is held annually. It's not inspiring innovation and discovery it's demanding a better version of what won the money last year, and the year before, and the year before, etc.
The trickier question to tackle here is whether or not passing a Turing Test can be equated to thinking. My initial reaction: no. I tend to side with the Chinese Room theory. That manipulating figures and essentially decoding, translating, and reguritating preformed responses (albeit in the right language) does not constitute thinking. In my opinion the Turing Test does little more than test the ability to read what I have said and come back with the correct response. The key word there is correct. In real conversation between humans there are millions of different possibilities for each statement and response. I don't feel that passing a Turing Test counts and really speaking as a human and therefore doesn't really count as thinking. It's too formulated and contrived. The computer generates no thoughts of it's own. Any thought is simply a manipulation of words and sentences given to it by a human.
No comments:
Post a Comment