“The ultimate hidden truth of the world is that it is something we make, and could just as easily make differently”
— David Graeber (1961-2020)
Link: https://t.co/Wa7pnhSKhh
@VictoriaVenken Paranthesized comment indicating what I meant by 'mass gap proof'; only a construction of the limiting distribution. I should have written existence.
Place your bets boys. My bets are Navier Stokes counterexample and Yang Mills mass gap proof (construction of the limiting distribution)
Of course, I don't know how trustworthy this account is
Alice: *does [stuff] to pay the bills*
Bob: *records Alice at work*
Alice: *continues to do [stuff] enjoying herself*
Bob: *uses data to machine replicate Alice's [stuff]*
Alice: "damn it [stuff] is now cheaper so I need to work harder"
Bob: "finally, I've successfully captured 90% of the [stuff] market"
Alice: "buggers now i need to find another job so I can continue do [stuff] in my spare time"
Bob: *makes a fortune*
Alice: :/
Bob: "Noooo! Nothing will save us from the incoming meaning crisis!! It's an existential catastrophe!!! Don't you see??? How will we have meaning in a world conquered by machines!!! This is terribleeeeeee"
@nihilunbounded People are not worried about loss of meaning because meaning is not something anyone else can extract from your work, meaning is something that comes out of you.
The economic value of one's work? Well that is definitely something the big AI labs are focused on extracting
Palantir CEO Karp: AI is dumping gasoline on a society that's already a fire. The people running the AI labs tell Americans their life is going to suck and they'll have no job.
Then they say: rely on me to give you money. By the way, I'm getting rich. 1/
This is silly on many dimensions, but the silliest dimension is that Machine God being better than me does not change my sense of superiority over my peers. :P
Like, Fabiano Caruana doesn't suddenly fall to his knees because a 900 ELO woodpusher can play chess better than him with an engine and slow time controls.
So rest assured that my fear is not one of suddenly being worse at producing papers than some shmuck with an LLM with no understanding of mathematics. I have many worries and fears for the community, but it's deeper than that.
Realized with a jolt that if LLMs were RL'd on chess enough it would probable evade all anti-cheating detection mechanisms on major online chess platforms like Chesscom and lichess and make the entire landscape of online chess uninhabitable.
Y'all do realize a world with advanced learning robots walking around is consistent with laws that prohibit any non-consensual use of human data right? It's just cheaper to boostrap off human data that corporations need not pay for seeing how powerless humans are vs capital.
@niloofar_mire@OpenAI it will not. 1.2B people use chatgpt man. revenue from ai labs is not that US/west concentrated. my friends who live in Bihar are on Instagram Reels
diffusion is slow when it is bottlenecked by humans understanding and integrating it. ai will understand and integrate itself
@willdepue@doomslide you need to get out of SF, sell all your assets and become a poacher in the eastern borders of Burundi. Nothing short of this will solve your problems.
@scottnarmstrong@KITP_UCSB This table from https://t.co/Ayk8NZb2ry has an answer :) they seem to have received another submission by OpenAI on November 4