Log in

No account? Create an account
06 January 2007 @ 01:27 pm
The Oz Project, back at CMU, was a fairly simple little AI deal -- the idea was that little spherical critters called something like "Woggles" would play with you. At the time it was still pretty nifty.

The folks that did that went on to found a company called Zoesis in the Boston area, and make a number of other demonstrations, and have some degree of commercial success. Their web site looks like they haven't done much since 1994, but I'll ignore that if you will.

One of their little demonstrations of a game-like thing with characters and motivations (their big deal) is called "The Pearl Demon and the Princess", about a captured princess and a demon captor. You can apparently download it and play around with it, but I haven't done that. I just cut straight to the chase and looked at the video of somebody else playing. It disturbed me. A lot. So I'm passing it on to you.


The link at the bottom says something like "download Mp4 movie of the latest demo". There's also a binary to play with it, but only for Windows.
(Anonymous) on January 7th, 2007 03:01 am (UTC)
What/Who/Where directed you back to the Woggles?
Noahangelbob on January 7th, 2007 03:20 am (UTC)
I was reading a paper about steering behavior (by a fellow named Reynolds) and it mentioned the Oz Project becoming Zoesis. So I hit Google to find out what they'd been up to.
r_transpose_pr_transpose_p on January 9th, 2007 03:07 am (UTC)
If you feel like exposing yourself to a bit too much math, you should know that the controls community is now starting to study swarming-related behavior as well.

is fairly seminal. For what its worth, the stochastic matrix theorem attributed to "Wolfowitz" was originally proven by Paul Wolfowitz's dad.

also look interesting, although I can't recall whether I've read them, so they might not be.


are fairly comprehensive, and have some neat problems, although all the wrestling with a proper framework for discussing multi-agent control may be only of interest to control theorists and theoretical roboticists.

is also a fun problem (for what its worth "rendezvous" is getting all the robots to have the same position, while "flocking" is getting all the robots to have the same direction of motion)
r_transpose_pr_transpose_p on January 9th, 2007 03:14 am (UTC)
Also, Herbert Levine,
has some interesting papers where he attempts to fit parameters in standard swarming models to observed data in actual bacterial swarms (apparently many bacteria move collectively as colonies)

Unfortunately, I can't find his papers online.
taoflaherty on January 7th, 2007 06:07 am (UTC)
The Woggles were impressive when I was 12. Lots of depressingly obsolete computer hardware was impressive when I was a kid. The Pearl Demon is a cute joke, but is any software even close to beating the Turing Test?
Noahangelbob on January 7th, 2007 06:13 am (UTC)
Nope, nobody's close to beating the Turing Test. Zoetope doesn't seem to be really trying in that direction.
taoflaherty on January 7th, 2007 05:37 pm (UTC)
Aiming to give characters personality, not so they're human but so they're interesting. Trying to decorate a room to make it pretty, not to make it indistinguishable from nature. Is their problem too hard for many others to be attacking, or just considered too irrelevent?
Noahangelbob on January 7th, 2007 06:19 pm (UTC)
The Turing test is a combination of hard and irrelevant - text interfaces are fairly passe, and speech interfaces are too far out to be yet credible.

Their problem, though, is probably just too poorly defined for others to be attacking it. They're discovering the difficulties in marketing it already... And it's hard to integrate it into the sort of game that would want it, further causing problems in marketing.
Katherine Donaldsonkatharos on January 8th, 2007 08:18 pm (UTC)
That's pretty cute! Dunno why anyone would want such a whiny princess though.
Noahangelbob on January 8th, 2007 08:32 pm (UTC)
To make the demon seem more attractive by comparison :-P