A former OpenAI intern is certain 2027 will bring a "century's worth of progress in a few months". What do you think he saw during his time at OpenAI to convince him of this?
>>17057089>What do you think he saw during his time at OpenAI to convince him of this?His stock fully vesting
>>17057089Nothing that threatens his future job prospect, meaning he still has to linkedin post on fucking twitter, which means that they're not replacing human jobs anytime soon.Granted, he's a fucking intern, so that might just be his best move regardless of how far along AI actually is.
the navier stokes issue means these breakthroughs won't be made public, so companies are going to be sitting on them like the us government was rumoured to with new tech for decades after world war 2
>>17057105they spent the last couple of years analyzing every letter that basedentists have typed into chatgpt and are preparing to steal their work?
>>17057089>breaking news: random dude on twitter said popular things with popular buzzwords
>>17057089He saw a smoothing function and a copy of the Central Limit Theorem.All errors average out.It’s really amazing that we have been equally wrong in opposite directions for so long.Thank you AI for showing us the way.
>>17057105OpenAI is sitting on a goldmine of bespoke artisanal forces which will make the oceans boil and the air turn solid.Wow.
>>17057107Tokens aren’t letters, Grasshopper.
>>17057105If they don't manage to solve another Millennium I'm betting this is gonna be the predominant cope.>nooo dude they actually solved everything but they wont reveal it brooo
>>17057089>a fucking furry
Wake me up when these AIs finally bring use something useful, like room-temp superconductors, cancer cures, or new life extension.
>>17057089the only rational reason to suspect a sudden increase in AI ability in the near future is if you think we're getting close to RSI
>>17057092Interns don't get RSUs
>>17057089i've been having psychic hypnogogic hallucinations telling me there's gonna be a crazy new model released for video generation within a four month time span and then shortly after an inferior, but still serviceable, model will be leaked to the public
>>17057089Very interesting indeed
>>17057412How exactly is a 200 page error-filled counter-example not life-changing for you?
>>17057089AI Moonboys?
>>17057089"Hmm, how will I be hired by the company I want to work at? Ah, yes, let me just say that they are god on Twitter."
>>17057089>A former OpenAI intern is certain 2027 will bring a "century's worth of progress in a few months". What do you think he saw during his time at OpenAI to convince him of this?Since he's pointing specifically to "mathematical superintelligence," lot of math, probably.
>>17057089I don't get it. How does he extrapolate it from the current trend of AI slop undermining and regressing every field it infects? Why would the magical electric golem do a 180 on this in Two More Weeks?
>acerfur
>>17057089Based
>>17057089>Math progressBooooring, who fucking cares about math problems. All useful mathematicians are physicists, engineers or in finance. Where is some actual useful progress like in bio or material sciences?
>>17057105>the us government>we painted our planes more black so radar can't spot them, it cost billions and is so sekrit and special>still gets btfo by Vietnam, Afghanistan and Iran, all poor shitholes>Iranians have known this for decades and use IR heat signatures instead of radar for targeting >fuck up dozens of military aircraft>their own allies with contractor jeet AA down three f-15s and then bully the pilots on friendly soillol, lmao even, burgers wouldn't last a day on the frontlines of Ukraine
