A company called AIseek got some press around 2006 for working on a dedicated AI processor called the Intia processor. I have yet to see this processor surface, but there is something on their site that strikes interest.

“AIseek’s Intia solution is integrated with the PhysX SDK from AGEIA”

Does this mean that the PhysX package that nVidia now has control over has integrated somewhere in it the ability to process AI?

Can we expect CUDA AI sometime in the near future?

CUDA already works very fine for AI, search the forum a bit, people have done neural networks, decision trees and other things.

While this is true of course, I think Aggressor Prime wanted to know if there’s a vendor made, ready to use API for AI (like PhysX for physics) that’s going to be implemented on driver level, like I’ve read PhysX is/will be on new nvidia cards.

PhysX will be supported on all CUDA capable hardware as far as I heard, so not only on new cards.

Good, I was hoping for that :)

That sounds interesting. Is there any ready to use API for AI at all? (hardware accelerated or not) AI sounds like something that is not easy to generalize to a small set of operations.

Never heard of one

Do you count chess also as AI? I have implemented a chess engine on CUDA.