The most prestigious law school admissions discussion board in the world.
|
 |
 
|
OpenAI tells NYT it may have solved a second millennium prize problem
| """''''"'"''"''''' | 09/10/26 |
| cannon | 09/10/26 |
| """''''"'"''"''''' | 09/10/26 |
| Jared Baumeister | 09/10/26 |
| Jared Baumeister | 09/10/26 |
| ,.,.,.,,,.,.,.,.., | 09/10/26 |
| """''''"'"''"''''' | 09/10/26 |
| \"\'\"\'\"\'\'\"\"\" | 09/10/26 |
Poast new message in this thread
Date: September 10th, 2026 12:45 PM Author: """''''"'"''"''''' OpenAI confirmed to the New York Times that they have made "substantial progress" on another Millennium Prize problem in the last five days, and are preparing to announce.
https://x.com/andrewcurran_/status/2098083604853342688?s=46
https://archive.is/20260910142243/https://www.nytimes.com/2026/09/10/science/tristan-buckmaster-openai-math-navier-stokes.html |
Date: September 10th, 2026 12:48 PM Author: cannon Someone named “Dr. Buckmaster” deserves to solve it (http://www.autoadmit.com/thread.php?thread_id=5902277&forum_id=2:#50126919) |
 |
Date: September 10th, 2026 2:38 PM Author: """''''"'"''"''''' that’s old news, this is a different Millennium Prize problem (http://www.autoadmit.com/thread.php?thread_id=5902277&forum_id=2:#50127321) |
|
Date: September 10th, 2026 2:45 PM Author: Jared Baumeister Is OpenAI saying he deserves no credit? Or is that just something you just made up? (http://www.autoadmit.com/thread.php?thread_id=5902277&forum_id=2:#50127341) |
|
Date: September 10th, 2026 7:37 PM Author: Jared Baumeister Anyone? (http://www.autoadmit.com/thread.php?thread_id=5902277&forum_id=2:#50128056) |
|
Date: September 10th, 2026 4:06 PM Author: ,.,.,.,,,.,.,.,.., He is my cousin. Not flame. (http://www.autoadmit.com/thread.php?thread_id=5902277&forum_id=2:#50127539) |
|
Date: September 10th, 2026 7:18 PM Author: """''''"'"''"''''' You the black sheep of the family? (http://www.autoadmit.com/thread.php?thread_id=5902277&forum_id=2:#50128009) |
|
Date: September 10th, 2026 7:48 PM Author: \"\'\"\'\"\'\'\"\"\" The transition to recurrent/looped LLMs likely allows them to rapidly scale up model capabilities without much additional training. The reason being that they can take a model with a certain number of looped layers and fine tune it for a larger number of loops without full retraining. Just stack more of the same layers, lightly retrain and model capability shoots up. This is mostly a curiosity (because of higher inference costs) but it allows the extremely rapid improvement from Astra to this model. Very deep looped LLMs could be useful if they want to automate AI research and kick off recursive self improvement soon. (http://www.autoadmit.com/thread.php?thread_id=5902277&forum_id=2:#50128073) |