An idea I once had
I don’t have the skills to set this up myself as it’s a large project and I don’t know how you’d go about setting up an opensource project on this scale. However.
Have you ever used the ALICE bot? I think it’s pretty amazing yet ridiculously simple. Literally anyone who knows basic English can make his own. And that’s my idea. A website like wikipedia based on user input that is moderated so that people all work together putting in a little time to make many different personality bots for a game. Then we get some open source coders to add the finishing touches to the game like the GUI. Sure it might not make money but hopefully it’d push the boundaries of games in the future. I don’t know. Good idea?
April 12th, 2008 at 8:13 pm
Somebody gonna whistle-blow…
April 13th, 2008 at 1:22 pm
Very good idea!
April 14th, 2008 at 5:00 pm
I’m not sure if it’s a good idea, but it’s a very unique one.
April 14th, 2008 at 7:11 pm
AIML is pretty cool, and I’ve often toyed with the idea of using it to simulate AI in a game or Live Person-style chat bot. One of the problems I always ran into was as the brain files got extremely large, the management of the patterns became more difficult. I started writing a tool to keep track of the hierarchical structure of the symbols, but I’ve since gotten sidetracked and I don’t know if I would have succeeded in developing what I wanted anyway. As a group project, you’d probably need such a tool and would also have to make sure that contributors understood creating the patterns, so as to avoid recursive or duplicate reductions.
Also the level of illiteracy in Internet communications makes the exact match nature of AIML cumbersome. An alternative would be to run a sort of translator and/or spell check on the input prior to performing the reductions. I wrote the spell checker, but again, it’s something I never got around to finishing.
It’s a really interesting concept though and I’ve always been surprised more people didn’t embrace it. Perhaps there are more advanced AI pattern languages nowadays?
April 15th, 2008 at 4:35 am
The reductions in AIML are probably one of the biggest problems with it. I don’t know if you could write an auto-reducer that just chooses the simplest reduction until somebody enters something which forces it to move into a more complex statement. I figured the massive number of people-hours available on the Internet would be the only thing that would make it work.
As to getting a more complex pattern system, it has to be really simple to use for the people actually entering the data to both save time and make it highly accessible.
Captain Blood is the only game I’ve actually seen that uses something similar. But that was just far too confusing to make it feel believable.
April 15th, 2008 at 12:57 pm
Indeed, cool. However, most folks are out there for the money, but I like the game idea for sure.
April 16th, 2008 at 10:16 pm
I don’t know how good of an idea it is, but I like it.