I bet they'll need to do RLHF
I was confused as to what relevant input would come from the Renfrewshire Local History Forum, or from a Rigid eLectrical Halogen-Free pipe. Until I scrolled further down the search results and found out about Reinforcement Learning from Human Feedback
Inflection AI (website, announcement) is a new startup founded by Mustafa Suleyman, Reid Hoffman, and Karén Simonyan. It appears they'll be doing something with language models. They're also looking to hire a safety team; I bet they'll need to do RLHF as well as toxicity filtering, etc.