I, too, wrote such a bot. It was more simple (ELiza-like with some patterns), could learn some stimulus-answer, but could also like or dislike people, according to what they said.
It is still on IRC, and is called Achille. http://francois.parmentier.free.fr/irc/achille_e .html It is not an intelligent one (no AI there, only random answers well-chosen). The bot does not reply everytime someone talks. That's an advantage. It talks only when it could give a correct answer. Indeed, some people thought it was a human for half an hour, but these were newbies on IRC (and I think in Computer Science), and the conversation had not such a great number of lines.
In this case, the users' motivation was important: the bot had the power in the channel, and they knew that, if it like them, they could be ops. So, even when they knew that this user was a bot, they tried to keep nice with it, talking with it, trying to prevent themselves from kicking the bot's friends, etc.
When the users are liked by the bot, they can teach the bot. So, its knowledge increases. And that's important too: it can then give unawaited answers, even to the bot's developer!
I plan to write a smarter bot, but lack time for that: ECTOR (based on my AI PhD). http://francois.parmentier.free.fr/ector/
But a really intelligent bot is something relatively far from us: even if we had the right algorithm, I think there would still lack the time for learning. Even a human being takes time for that (I mean, a baby takes a while to have a "normal" -ie well-formed- conversation). -- http://francois.parmentier.free.fr
Slashdot Mirror
I, too, wrote such a bot. It was more simple (ELiza-like with some patterns), could learn some stimulus-answer, but could also like or dislike people, according to what they said.
e .html
It is still on IRC, and is called Achille.
http://francois.parmentier.free.fr/irc/achille_
It is not an intelligent one (no AI there, only random answers well-chosen). The bot does not reply everytime someone talks. That's an advantage. It talks only when it could give a correct answer. Indeed, some people thought it was a human for half an hour, but these were newbies on IRC (and I think in Computer Science), and the conversation had not such a great number of lines.
In this case, the users' motivation was important: the bot had the power in the channel, and they knew that, if it like them, they could be ops. So, even when they knew that this user was a bot, they tried to keep nice with it, talking with it, trying to prevent themselves from kicking the bot's friends, etc.
When the users are liked by the bot, they can teach the bot. So, its knowledge increases. And that's important too: it can then give unawaited answers, even to the bot's developer!
I plan to write a smarter bot, but lack time for that: ECTOR (based on my AI PhD).
http://francois.parmentier.free.fr/ector/
But a really intelligent bot is something relatively far from us: even if we had the right algorithm, I think there would still lack the time for learning. Even a human being takes time for that (I mean, a baby takes a while to have a "normal" -ie well-formed- conversation).
--
http://francois.parmentier.free.fr