Exactly. The relentless "humanising" of LLMs really doesn't help. Its a noisy signal generated from either imprecise data, or not enough context to make a vali

Exactly. The relentless "humanising" of LLMs really doesn't help. Its a noisy signal generated from either imprecise data, or not enough context to make a valid token choice.