sorry, did any of you actually read metamorphosis of prime intellect before deciding to name your company that? prime intellect non-consensually converts the entire universe into a fake nanny-state playground where (according to the values of the MC) no humans can do anything that matters, so society degenerates into appalling disgusting hedonism, which the MC considers so bad she goes to great lengths to trick prime intellect into destroying itself, letting her restart civilization while keeping it permanently in the stone age so no one can invent AI again. why in god’s name would you name an AI company after this story???

People with a strong positive view on the Dvorak keyboard case study may disagree (or case studies with similar path dependency, legacy system inertia, local minima issues). I have managed to avoid acquiring a strong view on this so far.

after prompt engineering, context engineering, loop engineering, and graph engineering, the next big thing is... constraint engineering (i've always wondered how vocabulary emerges and becomes mainstream... i'd like to pre-empt the next hot phrase so i can look smart in the future 🫠)

My recommendation: don't talk to LLMs. Get them write code, do things for you, research stuff etc, but 1) don't talk to them 2) don't get them do your non-code writing.

I'm OK with that concept of alignment, and with calling this “law alignment” or something But taking “alignment” simpliciter to be the goal seems to have muddled the discussion