|
|
| |
Forum Overview
::
Are Games Art?
|
| |
|
by Senor Barborito 05/22/2003, 1:33pm PDT |
|
 |
|
 |
|
Entropy Stew wrote:
Senor Barborito wrote:
Yes, I've seen this. The solution to most AI-complete issues is to adapt a bottom-up approach rather than a top-down one.
Prove it. Go on.
Sure, here goes - humans do it. All human intelligence is based off a bottom-up approach (a massive highly adaptive neural network which you didn't use when typing the above). Gosh I hope the rest of this post is as easy!
Entropy Stew wrote:
Senor Barborito wrote:
An example of a bottom-up approach is genetic programs that result in the most efficient Alife.
GA != intelligence, and I don't know why people keep saying it does. Does the evolutionary process GA is based off of think? Fuck no. All it is is directed semi-random meanderings across the entire problem domain.
No shit genetic algorithms don't equate to intelligence, the idea is to try to evolve intelligence or at least intelligent behavior (most efficient path within domain) - I was simply providing them as an example of the bottom-up camp of AI (the smaller, newer one) which to absolutely clarify here would be the group of people attempting to arrive at intelligence via mimicking human behavior/biology/evolution. There wasn't much tolerance for them where I went to school, and it's still that way even within many AI departments today (MIT being a very large exception here).
Entropy Stew wrote:
Senor Barborito wrote:
neural networks
I'd tend towards calling this the "painful math" method. There are many network varieties, each with its own useful function (usually of a fuzzy nature).
Yes, I know, but without having yet defined the rather elaborate data structure topology, heuristics (buildings as opposed to park benches as opposed to hair brushes), or tolerances of our universe within which our proposed AI is to run amok it is impossible to determine which variety is best. Once this is done however, backpropagation (feeding human-defined universes that 'look good' into the network backwards) combined with many, many pass/fail trials for the network's output would result in a neural net capable of consistently outputting quality worlds.
Entropy Stew wrote:
Senor Barborito wrote:
Rodney Brooks over at MIT and he's pet project Cog would be the current flagship for this camp of AI.
OK, thats bottom-up. This guy is trying to teach intelligence to a humanoid robot through different forms of human-like interaction because he feels that human intelligence is a function of how we interact with our environment. Meh.
The other examples I gave you were just as bottom-up, I know what the guy is doing and he stands a hell of a lot more of a chance at producing something interesting than Lenat (1 in a bazillion as opposed to zero).
Entropy Stew wrote:
Senor Barborito wrote:
Top-down AI is logician's AI - heuristics, symbolic AI, etc. Doug Lenat's Cyc expert system is the penultimate example of this camp of AI.
Expert systems are good at specific problem domains, and telling you the logic behind the answers they give you. Nice, but they don't really learn on their own, and suck with generic scope.
No kidding. But a fusion of top-down and bottom-up is going to be required here - the 'work' is going to be in the bottom-up sector, not the top-down. That has as much to do with where grant money for AI research tends to go (thanks to conservatism in academia and top-down being the original approach) as it does with the relative strengths of each approach.
Entropy Stew wrote:
Senor Barborito wrote:
The trick is to use heuristic maps to set ranges within which highly flexible neural networks can run wild, using trial and error to teach them what works and what does not on a visual level.
That's even stupider than your Carmack-as-the-competent-implementor platform.
1. Your solution is actually a horribly disfigured genetic algorithm
No, it isn't. It's a neural network playing God within an extremely complicated heuristic topology. HEY GUESS WHAT, THAT'S EXACTLY WHAT GAME DESIGNERS ARE, TOO.
Only one doesn't have to simulate all or most or even a significant fraction of the game designer - one simply has to make a neural network which consistently outputs world-defining seed values that result in a gameworld most people would say is aesthetically competent if not 'stunningly beautiful.'
2. It would (sort of) work if the fitness evaluation function could JUDGE THE QUALITY OF THE GAMEWORLD
3. Judging the quality of the gameworld is AI-complete
4. The code to generate the world for each iteration would be the largest thing ever written if you wanted any sort of originality. 100% originality for this process would also be AI-complete.
-/ES/-
I'm certainly not suggesting 100% originality - I'm suggesting flash 10,000 towns at 100 people, and let each person grade the resulting town aesthetically. Similarly, backpropagate 1000 towns designed by 100 professional designers - the network will arrive at a set of weighted averages which produce results most people are somewhere between 'okay' to 'thrilled' with.
--SB |
|
 |
|
 |
|
|
|
| |
ARE GAMES ART? Avault investigates. by Fussbett 05/20/2003, 5:57pm PDT 
... by Lizard_King 05/20/2003, 6:04pm PDT
Re: ARE GAMES ART? Avault investigates. by Monty Cantsin 05/21/2003, 1:09am PDT
Re: ARE GAMES ART? Avault investigates. by Monty Cantsin 05/21/2003, 1:15am PDT
Shock and Awe by Entropy Stew 05/21/2003, 1:21am PDT
Re: Shock and Awe by ydrt 05/21/2003, 1:27am PDT
Maybe YOU should be the new about page!!!!! NT by Entropy Stew 05/21/2003, 1:31am PDT
also: HAHAHAHAHAHAHAHAHAHAHAHAHAHHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAH NT by Entropy Stew 05/21/2003, 1:31am PDT
We've got more about page jokes over here NT by GAY THREAD CONTAINMENT TEAM STATLER 05/21/2003, 3:01am PDT
Tell me more about We've got more about page jokes over here NT by Dr. Sbaitso 05/21/2003, 3:22am PDT
You know what fixes repetetive jokes? OLDER REPETETIVE JOKES! NT by Entropy Stew 05/21/2003, 3:34am PDT
BWA HA HA HA HA HAHAHAHAHAHAHAHAHA!!! NT by GAY THREAD CONTAINMENT TEAM WALDORF 05/21/2003, 4:30am PDT
I desperately hope I crash NT by The Forum SB 05/21/2003, 5:14am PDT
ATOR? NT by Y-DURT 05/21/2003, 5:16am PDT
Holy fuck. I never, ever, EVER thought I would see that. by Senor Barborito 05/21/2003, 3:44pm PDT
Re: Holy fuck. I never, ever, EVER thought I would see that. by Entropy Stew 05/21/2003, 4:50pm PDT
Pffft by Senor Barborito 05/21/2003, 6:02pm PDT
Re: Pffft by mark 05/21/2003, 8:44pm PDT
Ixnay by Senor Barborito 05/21/2003, 9:19pm PDT
"his" not "he's". Going to bed now. NT by Senor Barborito 05/21/2003, 9:20pm PDT
You just had to screw it up in the thread where you prove Monty wrong by Entropy Stew 05/22/2003, 2:58am PDT
Re: You just had to screw it up in the thread where you prove Monty wrong by mark 05/22/2003, 10:10am PDT
Re: You just had to screw it up in the thread where you prove Monty wrong by Entropy Stew 05/22/2003, 5:52pm PDT
Re: You just had to screw it up in the thread where you prove Monty wrong by Senor Barborito 05/22/2003, 7:17pm PDT
Re: You just had to screw it up in the thread where you prove Monty wrong by Monty Cantsin 05/22/2003, 8:05pm PDT
Um, no by Senor Barborito 05/22/2003, 8:26pm PDT
If I think you're already dead enough, will it happen? by Senor Barborito 05/22/2003, 1:33pm PDT
You're horribly, horribly confused on practically every level by Entropy Stew 05/22/2003, 5:31pm PDT
Your failure to even attempt being informed is making me hemorrhage. by Senor Barborito 05/22/2003, 8:23pm PDT
Just my artsy fartsy two cents... by Chairman Mao 05/22/2003, 8:55pm PDT
Re: Just my artsy fartsy two cents... by mark 05/22/2003, 9:07pm PDT
That is one of the funniest things I have ever seen... NT by Chairman Mao 05/22/2003, 9:12pm PDT
Well, here's the problem by Senor Barborito 05/22/2003, 9:10pm PDT
Re: Well, here's the problem by Chairman Mao 05/22/2003, 9:17pm PDT
Answer by Senor Barborito 05/22/2003, 11:05pm PDT
Re: ARE GAMES ART? Avault investigates. by Fussbett 05/21/2003, 6:37pm PDT
|
|
| |
|
| |
