Franz Laage

2023 / Jun 20 / 09:29 CEST

Some links that have been sitting in my draft notes for far too long now: “Simply explained: how does GPT work?” is a short, concise explanation of how LLMs work. It’s a very high-level take, explaining probabilistic approaches; no prior knowledge required. That whole probabilistic thing ties very well into “LLMs are good at playing you”. This one highlights the obvious limits of those models and warns not to anthropomorphize them. In “What are we going to do about ChatGPT”, Daniel Lemire argues that halting AI development for a given period wouldn't help anyone. It's not even enforceable. He gives a bit of history on similar initiatives.

← All atoms