Post

There’s a basic paradox of LLMs: for them to be effective at language, you throw everything you can at them. To make them effective at analysis and recommendation, you need curated data. Right now, we’re trying to wrestle sense out of the whole panoply of human intelligence and human stupidity.