Sam Altman says OpenAI is open to slowing cutting-edge AI
Quite a leap from "AI will destroy humanity" to "and the way to fix it is to give us more money" in 24 hours. If they want to babble about their destructive power as a marketing tool, put risk weights on them like we do for banks www.politico.com/news/2026/09...
David Dayen52"Oh yeah you can cause catastrophes? Then you need to carry $1 trillion in reserves for rebuilding purposes"
Ohma1Like crypto, they will want their bail out when the bottom falls out. I've seen this movie before several times, yup, I'm old
James Geluso1“Use AI for grid defense. There’s no way giving AI access to your grid will allow it to attack your grid on the way to destroying humanity.”
‘In a company-wide meeting this week, Altman told employees that OpenAI could potentially pace its AI development — perhaps in conjunction with several other AI labs, but that some may not agree to do so.‘ www.bloomberg.com/news/article...
dino-monster.bsky.social8They hit a wall on progress and really need to keep the investments rolling in.
M Martin2Sam Altman is open to throwing bullsh#t in every direction
Rlynner1This guy is dangerous AF. He's willing to do whatever the DoD wants, and now he's willing to put rails on other/commercial/individual development?? Way too many conflicts from a guy who seems way too conflicted depending on who is writing the check.
Virginia S. Ontario Possum1What if they paused for a few years to refine the product they have into more useful applications? What if the goal became something with concrete benefits? I'm all for investigative pure research, but these people are making products. Let them make better products for a while.
*OPENAI IS OPEN TO SLOWING CUTTING-EDGE AI, ALTMAN TELLS STAFF
Coach Finstock102Definitely not a scam. It is the least scammy product in the history of silicon Valley
ICE OUT State Fan5There’s a nonzero chance that they’re having trouble making improvements and that’s why they’re going “ohhhh it’s so dangerous” so they can use that as an excuse to explain slower advancement
Trantifa Prime3Ah yes, we’ve reached the point of the grift where people have clearly understood the “space data centers” were a grift from the moment the words left lips. So now, they kick the can of their failed society and enviro destroying product nobody wants… encourage regulation to hide the failure!
Steven Santos2Can the cutting-edge AI actually tell me how many r's are in strawberry?
Transient Sun2'we hit a roadblock and can't advance things anymore' but lying to make your company look better
G8erlaw682NONE OF OUR SHIT ACTUALLY WORKS SO WE ARE GOING TO SLOW HOW FAST OUR SHIT DOESN'T WORK SO NO ONE FINDS OUT OUR SHIT DOESN'T WORK.
🏳️🌈Ash❤️🔥Cinder🏳️🌈2This shit reminds me of how the director of Foodfight infamously gave his animators instructions like “Make it more awesome,” or “I need that 35% better.”
A pleasant side effect of OpenAI slowing down its cycle of spending billions training a new frontier model every few months is that it should improve its economics. Training is a massive cost center. Selling inference (AI usage) to paying customers is increasingly profitable.
Dominic Hopton3One read on this is cover for getting the containment systems sorted out, so they slow everyone and then ”we fixed it” full throttle in 12 months when they fix it. Or they realize they can milk the current state for a while and don’t want anyone else to 1-up them.
Brian Drell1Maybe they’re figuring out that a recursively self-improving autocomplete algorithm still gives you an autocomplete algorithm. But I suspect this insight continues to elude them.
