AI should only run as fast as we can catch up

Nondeterministic isn't the right word because LLM outputs are deterministic and the tokens created from those outputs can also be deterministic.

I agree that non-deterministic isn't the right word, because that's not the property we care about, but unless I'm strongly missing something LLM outputs are very much non-deterministic, both during the inference itself and when projecting the embeddings back into tokens.

I agree it isn't the main property we care about, we care about reliability.

But at least in its theoretical construction the LLM should be deterministic. It outputs a fixed probability distribution across tokens with no rng involvement.

We then sample from that fixed distribution non-deterministically for better performance or we use greedy decoding and get slightly worse performance in exchange for full determinism.

Happy to be corrected if I am wrong about something.