Guys,
Thanks for the input. I really appreciate it.
[:I]To show how green I am, I had envisioned a bot that would operate just as a real employee would. They would know certain things to be fact (such as the product is always good -lol) but learn other things (like the customer's name, issues, and preferences). Perhaps Turing should have added knowing when to lie to his test for artificial intelligence. :-) But then, I guess he did when he said it would fool you into thinking it was human.
I have until May of 08 to complete my project, and this exactly why I wanted to start the proposal process early. I'm still very ignorant about the topic and I don't want to propose something that will only make me nuts in the long run.
A certain amount of research is required for the project and I'm actually thinking I may be on to something with all this. It is not necessary for me to prove my thesis, just merely come to a conclusion. I have downloaded both answerPad and Verbot and will pick up Hal today at work when I have highspeed. What I'm thinking is to train and compare perhaps three different bots and see which performs the best at interaction with my data and my 'customer'. I could also look at ways to tweak the code, scripts, and data to arrive at optimal performance (in an attempt to get a bot that wouldn't be fired it's first day if it were human).
I'm still thinking that Hal is probably not a good candidate for inclusion in my project because the code is not open source. I would need to understand why he performs the way he does in order to provide any useful information to my conclusion. Also, my sponsoring prof is a bit of an AI enthusiast, so he would probably want to see the code.
I hope that I will still be welcome in this forum. I find the discussion fascinating, and the people warm, friendly, and passionate about the topic.
Gratefully yours,
Angela