i'm not saying you should. i'm just saying you could.

i still have strong feelings about this book and the way it argues we should think about AI, snake oil, regulatory interventions, etc... - and acute scorn would be warranted - but i'll settle for not giving it the oxygen of a direct engagement in the form of something like a blog post or a thread.

i'm almost certain we'll do it again. i'm just not sure precisely when, but we'll promote it more lol

this is why SB 1047 is so fucking maddening. there's a definition of what constitutes a dangerous AI system that's predicated on how many integer calculations or floating point numbers are involved like what the fuck are you even talking about - oh my god it's happening again i'm sorry, i'm done.

AI is an overarching political / ideological project to displace and relocate autonomy into the hands of technocrats and other tech elite. fixating on the technical stack is stupid; even AI researchers played the field and dabbled in expert systems. Prolog is the "first AI programming language".

i'll just say that people can talk about a constellation of phenomena and we can all understand and appreciate the coherent meaningfulness of calling it "AI". CS people don't get to monopolize language - not even technical jargon. what a weirdly meta conversation to be having by the way

in case anyone's anxious: faceID will work again as soon as you enter your passcode. it's a very temporary measure, but i prefer this approach over disabling faceID at a protest because if you're in public constantly entering your passcode, someone could see what you're typing in, which is very bad.

god why

(But of course there is no contradiction that the most robust defenders of the idea that "LLMs work just like people" aren't super concerned about the value of precise language)

this is amazing.