The most prestigious law school admissions discussion board in the world.
|
 |
 
|
So “AI” has basically hit a wall and isn’t going anywhere?
| .,.,....,.,.,.,.,,.,.,.,.,.,.,.,.,..,,,.,., | 09/12/26 |
| The Penis | 09/12/26 |
| .,.,....,.,.,.,.,,.,.,.,.,.,.,.,.,..,,,.,., | 09/12/26 |
| The Penis | 09/12/26 |
| \"\'\"\'\"\'\'\"\"\" | 09/12/26 |
| The Penis | 09/12/26 |
| Jared Baumeister | 09/12/26 |
| Consuela | 09/12/26 |
| ~~(> ' ' )> | 09/12/26 |
Poast new message in this thread
Date: September 12th, 2026 5:51 PM Author: .,.,....,.,.,.,.,,.,.,.,.,.,.,.,.,..,,,.,., Sad tbh, everyone was so excited
https://x.com/drewhamlett/status/2098828189204115870?s=46 |
Date: September 12th, 2026 5:55 PM Author: The Penis No the jumps are getting bigger than ever. Right after Open AI released gpt 6 they had an internal model that is significantly better solve a millennial problem. (http://www.autoadmit.com/thread.php?thread_id=5903308&forum_id=2],#50132354) |
 |
Date: September 12th, 2026 5:57 PM Author: .,.,....,.,.,.,.,,.,.,.,.,.,.,.,.,..,,,.,., I saw that, ChatGPT stole the math proof from an IRL mathematician. This shit is all flame. |
|
Date: September 12th, 2026 6:04 PM Author: The Penis Yeah people think it stole the proofs but its all flame people are gullible. It was chat gpt that solved it not the dumb lying human. (http://www.autoadmit.com/thread.php?thread_id=5903308&forum_id=2],#50132382) |
|
Date: September 12th, 2026 6:13 PM Author: \"\'\"\'\"\'\'\"\"\" It’s not clear what happened with the millennium prize problem. Even in the worst case scenario where they relied on data trained from chat sessions with the researchers, GPT 5.6 did a substantial amount of the research. In any case, the Astra results that were released previously has no questionable authorship and were all significant results - several plausibly Fields medal worthy.
The models are still substantially improving on verifiable tasks and there is no evidence of a slowdown. The SWE improvement is substantial enough that agentic models are increasingly capable of autonomous AI research, so the concerns about significantly better models in a year or two starting recursive self improvement are plausible. The researchers are actually concerned. This isn’t a marketing ploy. |
|
Date: September 12th, 2026 6:18 PM Author: The Penis The millennium prize one wasn't gpt 5.6 it was an internal model that is a huge leap from astra at math. The Open AI researchers published their paper publicly it is radically different from what the guy thinking his ideas got stolen was doing. Even in the unlikely chance his prompts did get trained on, no way that would be in any way meaningful toward getting the result they did. The mathematician was using gpt 5.6, and they were using a way better model and spent millions in compute on it. (http://www.autoadmit.com/thread.php?thread_id=5903308&forum_id=2],#50132451) |
|
Date: September 12th, 2026 5:58 PM Author: Jared Baumeister the LLMs are going open weight, but whatever you build with them is still yours. It's 180 (http://www.autoadmit.com/thread.php?thread_id=5903308&forum_id=2],#50132359) |
|
Date: September 12th, 2026 6:15 PM Author: Consuela like your wife haha (http://www.autoadmit.com/thread.php?thread_id=5903308&forum_id=2],#50132437) |
|
Date: September 12th, 2026 6:30 PM Author: ~~(> ' ' )> ChatGPT's image creation is getting better. (http://www.autoadmit.com/thread.php?thread_id=5903308&forum_id=2],#50132484) |