OpenAI researchers recently claimed a major math breakthrough on X, but quickly walked it back after criticism from the community, including Deepmind CEO Demis Hassabis, who called out the sloppy communication.
It started with a now-deleted tweet from OpenAI manager Kevin Weil, who wrote that GPT-5 had "found solutions to 10 (!) previously unsolved Erdős problems" and made progress on eleven more. He described these problems as "open for decades." Other OpenAI researchers echoed the claim.
The wording made it sound like GPT-5 had independently produced mathematical proofs for tough number theory questions - a potential scientific breakthrough and a sign that generative AI could uncover unknown solutions, showing its ability to drive novel research and open the door to major advances.
What’s the salary of the top AI researcher, trainer, etc? Fake speech and Counterspeech, redux:
Terrace Garden, first floor: OpenAI tweets. “GPT-5 solved 10 (!) unsolved math problems!” Neon lights flash, investors glance up, media sirens wail. Some people believe in the clean, gentle ritual!
Middle floor terrace garden: researchers echo it, press releases gleam like mirrored glass, the lobby fills with astonished visitors. Numbers obfuscate — 10 problems, 11 more in progress — the Manhattan skyline is becoming more dense and population is what?
Upper floors: scrutiny arrives. Someone points out the wobbler and wobbly beams wobbling : lackluster communication, over extending and inflating the background and progress like retrocausality . . .
Suddenly, the shiny becomes more shinier and the tower tilts, the tweet disappears, glass panes of hype crack a little more like a cracked foundational arch acting as a bridge to the top.
Penthouse: reality check. GPT-5 is powerful, but skyscrapers built on miscommunication, misleadings, misguidance, confusion and mistakes are fragile, not robust nor anti fragile.
True breakthroughs need mighty and restored foundations like retrofitted superstructures, not more press releases. The Manhattan I once knew remains impressive, but let’s not confuse reflections in tinted windows with the sun on the streets during the golden hour.
Now, next time these - to speak with LeCun - GPTards tweet something, please please please do us all a favor and think: are they just fooling themselves?
Cuz they lie to themselves all the damn time. OpenAI is what you get when you convert your echo chamber into a giant pressure cooker.
GPTards
tweet something, please please please do us all a favor andthink
: are they just fooling themselves?