An Architecture for Conversation
In this article, “Fooled by a chatbot” the author discusses the fact that chatbots remain an interesting study for AI because people aren’t always using the maximum of their intelligence and that chatbots can pass a Turing Test in which the participants aren’t “on guard”. While I agree with this statement, I’d like to offer a separate reason why explorations of simulated intelligence or chatology (heh heh) are more than just trivia or entertainment. Namely, they represent a potential breakthrough in user interface design. One of the shortcomings of the graphics user interface is that it forces users into a serial method of working with data “objects”. It is very difficult for a user to remove the “b” from 25 file names all at once in a GUI system. And most other methods require knowledge at the command line level or enough programming savvy to write and execute some kind of script. Both of these options are little more than swtiching from a GUI to a LUI or linguistic user interface. The problem is that a LUI generally requires that the user speak a very specific language with very strict grammar and syntax. Chatbots are opening up the potential for LUIs by making an attempt to understand English.
Take the above user problem and imagine how a user with an english speaking bot in his computer could accomplish what they want. Even if the “English” that the bot spoke were slightly restricted and simple, there are a number of tasks that are far simpler as a result. I, for instance, have been using SmarterChild for quite some time now and it has proven to be invaluable at some tasks. This is primarily for three reasons. One is that my chat window is virtually open all the time. The second reason is that I can use (and remember) the simple commands that it acccepts (like “spell presient” and it will figure out what I mean). Three is that it remembers things about me and makes other tasks easier. I don’t look up movie times on the web anymore. I just instant message SmarterChild: “I want to see a movie” — he knows the drill.
[1/24/05 – note: No sooner had I posted this entry than I found a similar article over at o’reilly that asks the question “Are chabots the command line of the future?”]