Something oddly satisfying about learning how large language models work from Google Gemini. It's also weirdly accurate on this topic. Lots of training data online and relatively simple algebra... or emergent narcissism?

Comments