Slashdot Mirror
A Borg-like Artificial Intelligence For Lionhead's New Game
cybaea writes: "The creator of Black & White is experimenting with new work on group minds - but unlike the Borg, the characters in the new game are already descending into bar brawls,
reports ZDNet UK, quoting Richard Evans (famous for the AI engine in Black & White). My favourite quote: '[AI] Characters [in the game] even have the ability to dynamically create their own language, constructing simple sentences on a word by word basis.'"
I'll believe it when I see it.
AI constructs may evolve from things like this, but they'll need a home on the internet, in order to have lifetimes long enough to become really sentient. humans require years to become intellectually complex, from preexisting instructions worked out over millions of years. When these constructs have a semi-stable environment, modification, and competition, it should be just a matter of time....
"Characters even have the ability to dynamically create their own language, constructing simple sentences on a word by word basis."
o graphyItem.php3?reference=steels%3A99f)
I wonder whether this means creating new words and then constructing sentences using the new words, or if the characters will be given a lexicon and a grammar and will produce sentences using them.
The first case is quite time consuming. Many iterations of language development "games" are required to produce a common language. Also most of the language development processes that have been proposed only produce a limited subset of the syntactic categories. There would also be the problem of the person playing the game being unable to understand the AI characters. (For information on language development see http://www.csl.sony.fr/General/Publications/Bibli
In the second case, would the characters be able to produce syntactically correct sentences? The 'goodness' of the sentences would depend, I guess, on the size of the lexicon and the complexity of the grammar rules. However producing complex sentences would make it more difficult for other characters to understand them, due to the difficulties of parsing a rich language. I just hope it doesn't end up being a (subject, object verb) language with no real syntax.
I will be interested to see just how this turns out.
I mean, a group of minds is not the same thing as a group mind.
Getting into a barfight dosn't seem like something the borg would do.
autopr0n is like, down and stuff.
Come in please.
We really think your technology is great. Your graphics are excellent. Your scalable-terrain engine knocked our socks off. Your physics engine is amazing. Your AI code is already quite remarkable. Your simulation of a believable, detailed fantasy world is outstanding.
But Lionhead, we have a problem.
Black and White just wasn't fun to play.
Once we were done being amazed at all the features and gasping at the technology - the game just wasn't very good. It didn't engage. We weren't motivated to continue. It just got boring. Sorry - no-one wanted it to be great more than us, but in the final analysis it just wasn't.
You guys are great. You plainly love what you do, and create high-quality product. We're grateful for your dedication. But please - make the next game fun first, then add in the AI, the nice graphics, the believable simulation. We appreciate that fun is hard to describe, hard to measure, hard to design or schedule or test. But it's important. It's only fun that separates a game from a fishtank.
Thanks for listening. Earth out.
## W.Finlay McWalter ## http://www.mcwalter.org ##
Sure, this might be a good sign for general programming evolution, but it didn't prevent the first Black & White from fizzling due to virtually zero replay value and plenty of bugs. PC Gamer gave it an Editor's Choice back in about May 2001, but that was mere weeks after the game was released; months later, player opinion of the game plunged. Black & White, Lionhead, and Peter Molyneux became the butt of many jokes. It wasn't like Id Software's games, where a great engine was held back by a vacuous storyline. The engine was buggy, the principle was weak, and even the AI had problems. People asked themselves whether the developers at Lionhead had played their own game through to the end. Personally, I've learned my lesson after purchasing stinkers like Red Faction purely on the speculations spewed out by "gaming sites", only to find out that the game wasn't worth 1/4th its release MSRP. And it seems that the good games are taking forever to develop since the developers are actually playtesting them and making sure they don't mess up during development. There's going to be a long stretch of time before the good games get released, while the discount devhouses pump out half-developed games by the truckload.
"Ancillary does not mean you get to rule the world." --U.S. Circuit Judge Harry Edwards, speaking to the FCC's lawyer
The whole Enterprize series is a screwup!
This must be the latest way to get free publicity. How many articles in the last few months said something to the effective of "IT's ALIVE!". Robots that have "learned" how to fly all by theirselves (yea right), game AI that is developing it's own language (yea right). Apparently it works, Slashdot falls for it everytime and publishes the story.
I'm Rick James with mod points biatch!
take a course at your university about neural
networks and you'll see they are nothing great.
at the best they can learn about what you want
to teach to them, and that's all. usually
they'll just learn a very erroneus version
of your teachings. building a neural network
is pure heuristics. adding a neural network
to a game doesn't add any character to it.
it just learns to mimic a little. i don't know
what's the fuss about neural networks saving
your game play OR THE UNIVERSE, but i think
nn is just a (bad) tool. you'll need a lot
more than that to make a good game. and i'm sure
you can make a really good if not the best without
any nn bs.