AI safety plan people asked for:

we'll get all the smartest people
we'll lock them in the basement.
when we make the smart ai,
we're going to lock it in a castle.
the castle will be defended by the government.
other people's castles get the airstrikes

AI we got:
the kids will use AI to fake learning
the teachers will use AI to fake grading / teaching
the White House will use AI to make punitive economic policy
HHS will use AI to make claims and punitive restrictions
MechaHitler
the DOD will buy AI after MechaHitler
the World will invest in AI
the US economy will shrink except for AI and elder care
the AI will burn gas and coal.
batteries will be tariffed.
nuclear will be vibes.
the AI czar will use AI to make AI summaries
AI financial advisors will plug in to produce data.
maybe soon they'll transact directly

The doomers weren't right exactly. Foom didn't happen (Couldn't? I argue).

Well before AI could threaten to be a superhuman AI researcher and self improve, humans are tempted to jam it in everywhere, and assume it works, before it really does. YOLO for yo LLM

surely this becomes iterated disaster (we're in it) but like, fortunately a survivable one. but that means smart people should be trying to mitigate it now. there will be a butlerian backlash, you already see this

https://t.co/Sb1w8AKWCW