On: How Meta trains large language models at scale

[vineyardmike]

Because people generally don’t use TPUs outside of Google. The tooling is different, the access is metered through GCP, etc.

Nvidia is in a vaguely unique position in that their products have great tooling support and few companies sell silicon at their scale.

[refulgentis → vineyardmike]

Correct, I'm pointing out politely that's in conflict with the person I'm replying to.