, aq

-Power strip quote: "WARNING: In order to properly protect your phone/fax/modem follow these steps:"-

This past week was a very interesting one dreamwise, seeing as I'm now finally getting abundant amounts of sleep. Let's see. I had a false awakening dream that sort of evolved into musings about the "dream paralysis" phenomenon, and how it could be used for nefarious purposes. I had another dream in which I discovered that fairies wear knitted dresses and also spent a long time thinking about those clocks that wind themselves up by fluctuations in atmospheric pressure. The next night, I had a dream where I died (once again disproving the idea that you die in real life if you die in a dream), yet mysteriously ended up in merely a whitened version of my childhood house, which cannot possibly be accurate. Finally, last night I dreamt that I was driving to an airport using arrow keys somehow, but as I was in third person perspective as usual, I was having trouble controlling the car smoothly due to the fact that the "camera" wasn't tracking it very well over hills. I made it to the parking area just fine, but a policeman asked for my license shortly after I stopped.

In my patterings around the Internet this past week, I happened upon a rather well-thought-out chatterbot. A chatterbot is a program that humans can chat with (often online) that aims to replicate human speech patterns in such a way as to be indistinguishable from a REAL person chatting. So far this has not been fully accomplished, as can be deduced by the fact that no AI system has yet passed the Turing test.

However, chatbots can be awfully fun to mess with, especially when the coders obviously did not think many conversations through at all. Many of the bots have ridiculously simple reply algorithms that lead them to respond to only one key word in your statement, even if the sentence as a whole had nothing to do with it. For instance, if the chatbot says something confusing, you may say "I'm sorry, I don't understand" and the bot will respond with "Apology accepted," or something equally inappropriate.

Anyway, the thrill of these chatbots is usually short-lived, as the conversation gradually degrades to total confusion on both sides. That is, until I found Jabberwacky. I'm not sure what makes the actual program different from the other bots, but I've found that I can have a much more glitch-free (though by no means perfect) conversation with Jab. The only thing that's consistently a problem with this particular chatbot is how WELL he imitates humans. Meaning that he will often accuse me of being a robot, while staunchly defending his own imagined humanity, much as the people conversing with him probably do. Also, he's a lot more abrupt and curt than a normal person, given that people are not quite so polite to something they know is not actually offended by lack of courtesy.

This of course made me think of the perfect idea for a chatbot. Instead of making it obvious to the humans that they are conversing with a bot, introduce the chatbot to a normal chat room (where people are told they are being observed but are randomly matched with either people or bot) under different names. Have it watch the human-to-human (e.g. normal) conversations a lot and only occasionally be paired with a human, without announcing the fact that he is any different. Obviously at the beginning he would be fairly awkward and noticeable, but if he always assumed a different name, he might eventually be able to talk for quite some time with someone before they started treating him as a robot.

This scheme would have the added advantage of making people come back again and again because they'd never quite be sure whether they got a chance to speak with the bot. It would make an interesting hook.

Reality check. Brought to you by Blogbot 4.0