Re:We're in the brute force phase of AI – once it ends, dema...
by kolinko on Tuesday September 10, 2024 @10:21AM
(#41501076)
You can get away from that if you constrain it to a specific type of models (say attention based).
Re:We're in the brute force phase of AI – once it ends, dema...
by adastra22 on Tuesday September 10, 2024 @10:35AM
(#41501226)
You don’t need general programmability for AI inference.
Re:We're in the brute force phase of AI – once it ends, dema...
by glitchc on Tuesday September 10, 2024 @11:59AM
(#41502173)
The money's in the training, not the inference.
If you look at Apple and Google, they already have their own hardware for inference in their smartphones. They don't need NVidia for that.
Re:We're in the brute force phase of AI – once it ends, dema...
by mupuff1234 on Tuesday September 10, 2024 @12:08PM
(#41502302)
Apple and Google use TPUs for training.
https://www.cnbc.com/2024/07/29/apple-says-its-ai-models-wer...
Re:We're in the brute force phase of AI – once it ends, dema...
by glitchc on Tuesday September 10, 2024 @12:21PM
(#41502495)
Hmmm, that's worse for NVidia.
And, as it happens, that's exactly what NVidia's done with the H100: https://developer.nvidia.com/blog/nvidia-hopper-architecture...
It still needs to be programmable though. Can't get away from that.