CPE/CSC 480-F06 Artificial Intelligence Lab Exercise 1: Chatbots

One of the first AI programs that attracted the attention of a wider audience was a program called Eliza, developed by Joseph Weizenbaum in the sixties. Eliza was modeled after a technique used in psychotherapy where the therapist asks the clients questions by referring to keywords mentioned by the patient. While nowadays hardly anybody would be fooled by Eliza into thinking that the program "understands" them, at the time, some people developed strong emotional attachments to it, and a few people went so far as to claim that programs like Eliza could replace therapists under certain circumstances.

In the meantime, many programs that are much more sophisticated than Eliza have been developed, and quite a few of them are available on the World Wide Web, or in chatrooms. They are frequently called "chatbots", "chatterbots", or "chatterboxes".

The links below point to information about chatbots. A Web search on Google or another search engine can easily lead you to many more pointers.

• An overview of some of the early chatbots, including Eliza
• The entry for "chatterbots" in the Open Directory Web site, with lots of links to chatbots
• The Turing Test entry in the Open Directory
• The home page for the Loebner Prize, a competition inspired by the Turing Test

In this exercise, your task is to evaluate two of these chatbots with respect to at least five criteria for intelligence. These criteria don't need to be defined very carefully at this point; part of the exercise is to examine the usefulness of such criteria. For each aspect, briefly describe why you believe that the chatbot does or does not exhibit relevant traits. At the end, describe your overall impression of the chatbot, and how "intelligent" it appears.

You can use the table below for this exercise, or create your own.