Superalignment

An LLM is trained to continue text plausibly. That single ability turns out to cover summarizing, translating, writing code, extracting data and holding a conversation. What it does not include is any guarantee of being right. Plausible and correct are different targets, and the model is optimized for the first one.

Why it matters

It explains the characteristic failure: fluent, confident, wrong.