6 ms·
Are deep neural networks really that widely applicable that it's profitable to design custom chips for them? What about other models of AI that involve, say, di
by brookhaven_dude 7y ago
Are deep neural networks really that widely applicable that it's profitable to design custom chips for them? What about other models of AI that involve, say, discrete math or graph search?
- Q6T46nT668w6i3m 7y agoI don’t know whether it’ll be profitable, but MATMUL, for example, is useful for a variety of programs beyond propagation. My guess is most of this stuff will be packaged (e.g. Apple’s “neural engine” on their A-series SoCs).
- modeless 7y agoYes. They are far beyond any other AI technique in speech recognition, speech synthesis, translation, OCR, object recognition, playing Go, and many other diverse tasks. And their performance continues to increase with added computing power with no limit that we've seen yet, so custom hardware improves results.
- streetcat1 7y agoAlas, you do not usually train models from scratch. I think that transfer learning will dominate, and it does not need this power.