Hacker News Reader

> Splitting TPUs into dedicated training vs inference chips feels like an admission that the bottleneck has shifted from FLOPs to memory bandwidth + latency.

With the expected scale of inference, it makes cost sense to make dedicated hardware for each task if the workloads are even slightly different. Probably similar to the video decoding chips in TVs not being very cheap/efficient compared to chips capable of encoding video.

> admission that the bottleneck has shifted

There's no admission - this has always been known.

I think the first two paragraphs of the post are exactly saying that the bottleneck is memory... Long contexts, bigger but less flop-intensive models (moe's).

The funny thing about scaling laws is that as soon as they were known, the whole objective became learning how to break them - bending the curve, at least. They provided an incredibly useful target, but 'law' was a bit too strong a word.