For those of you haven't heard, A.L.I.C.E is an intelligent chat bot which has won awards for it's ability to carry on a conversation online. This dialogue on nik - miscellaneous ramblings demonstrates what happens if you have one instance of A.L.I.C.E talk to the other. The original Slashdot post where I found the reference to this site points out that "results show that the bot is not really that intelligent, and relies more on human input."
Agreed. The dialogue does break down in points, particularly at the end. But at some points, the disconcerting directions the conversation (can I call it that?) takes is reminiscent of online group conversations with students fluent in IM, and, indeed, of some real life conversations.
Re: A.L.I.C.E Bot-Off
Perhaps this is nitpicky, but are most conversation bots given female names? I only ask because of the way in which these bots are doomed to always respond rather than initiate; to appear interested in everything I say rather than have their own interests. Reminds me of the one-sided conversations one of my female professors has told me about often having with her male colleagues. Think of all the bots for playing chess: Hal, Zarkov, Deep Fritz--all male.
If the programmers were familiar with classical rhetoric, perhaps they could call their chat bot Ion. ;)
See what A.L.I.C.E. thinks!
Re: are most conversation bots given female names?
looks like maybe not
Re: A.L.I.C.E Bot-Off
I remember the original Liza program. I had it on my C-64; not too intelligent then, but still lots of fun. Later models could simulate typing mistakes and other human flaws.
I always felt that Artificial Intelligence should be renamed to Artificial Stupidity, or maybe Artificial Stupidity Simulator. We suspect the perfect and even purchase clothing now that appears to damaged or worn.
Re: Artificial Intelligence should be renamed
Either that, or educate people about what AI is really about. When I took AI in the mid 80's, the principle behind AI studies is not that computer scientists are necesarily always trying to produce programs which yield the right answers; rather, that the principle of regular algorithm design is to produce a program which always yields a correct answer/post-conditions to a particular range of inputs/pre-conditions. On the other hand, AI attempts to modify some processes of the human brain in reconition that the brain is often more efficient than a machine, although the algrothims based upon human mental processes do not always yield a correct answer/satisfy the post-conditions. AI's, then, will always be limited and sometimes ineffective, just as we are :)