The Grok thing is low-key fascinating. It shows how someone like Musk can fuck up the algorithm to present irrelevant info to a mass audience, while also showing that so long as a LLM depends on mass sources for its info, it can do so in a way that essentially negates the intention of the fuck up.
May 15, 2025 20:10It suggests that for the right to fully use "AI" for its purposes it needs to fund enough non-AI writing to contradict the factual writing that disproves their points. If a Musk-owned LLM ordered to refer to "white genocide" in every answer can't find evidence of a white genocide, that says a lot.
