A.I. Microchip PCI Card

Started by
2 comments, last by Prozak 22 years, 7 months ago
I''m very sure someone has to have had this idea before me, if so, what are they waiting for? My idea is simple. Neural Networks and other methods for creating AI agents, have all been defined. A company could simply create a chip with arround 1 million neurons, and we the programmers, would fill that chip, and make the axional conexions. Then we would say "render", and the chip would process its data, sending us the output results to a temporay memmory buffer, with which we could decide how to move our alien monster that is creeping in our mothership in our game... ok, I think u guys have the gneral idea, and no, I don''t believe we should stick with NNets only. We could go several ways. If AI is one of the biggest piece of the pie, in processing time, then with hardware support, imagine what we could do... ok guys, waiting for your input on this one, Hugo Ferreira UniteK Future "Concentrate, and you can see it. If you see it, then it is possible. If it is possible, you can Achieve it."
Advertisement
It's a great idea, Ive thought about it several times myself - especially since it was mentioned in Edge100.

The only problem I can see in it is convincing the general games public to buy one - many people today complain about AI but will only upgrade their graphics card. Many so called "Enthusiasts" build cruddy systems but put excellent graphics cards in it, hoping for awesome graphics and ignoring the importance of a fast hard drive, processor and ram. These people are not going to invest in an AI card, which is a real shame - because the increase in speed would be phenominal.

I recently heard that John Carmack said the Geforce3 is as big a step forward in games as the voodoo was... personally I think the creation of an AI card would hop-skip both. Especially with the latest games coming out being more & more AI intensive.

Neway Im gonna stop now,
Crash,




We Must Move Forwards NOT Backwards, Sideways NOT Forwards And Always Twirling Twirling Towards Success

Edited by - crash on September 10, 2001 12:20:13 PM
"We Must Move Forwards NOT Backwards, Sideways NOT Forwards And Always Twirling Twirling Towards Success." - 2000"If You Keep Looking Forward Your Gonna End Up Looking Backwards At Yourself Running Sideways!" - 2001
But who uses neural nets in an actual game? You say you''re going do more than neural nets, but there''s no "general" approach to AI that can work for lots of games. Developers don''t want to limit their creativity by using a piece of hardware that only works for cerain types of AI. Heck, there''s lots of people who feel that 3D accelerator cards don''t give them enough freedom - imaging saying "you must use neural nets in your game to make it hardware accelerated".

If you can find a general approach to AI then it would be a great idea. But I think that''s a long way off. What I think we really need is for everyone to have (at least) dual processor systems, so we can dedicate one whole processor to AI.

codeka.com - Just click it.
Who did full 3D geometric transform and lighting in real-time in 1992? (yes yes, SGI, I know)

I think Dean''s on the right track. You can put a CPU on a PCI card, give it a DMA channel, and away you go, a multiprocessor system upgrade on a card. They exist now for doing RT system control in concert with Windows.

The whole AI thing is probably better solved by using multiple PCs and a network (I guess they said that about 3D graphics processing back in 1992...).

"Turn your spare computer parts into a cunning enemy!"(tm)

Magmai Kai Holmlor
- Not For Rent
- The trade-off between price and quality does not exist in Japan. Rather, the idea that high quality brings on cost reduction is widely accepted.-- Tajima & Matsubara

This topic is closed to new replies.

Advertisement