OneManyNone

6 ms·

1. 1.▲
   by OneManyNone 7d ago

   “ On Tuesday, September 1, we heard rumors that two Millennium Prize problems had been resolved. Inspired by these rumors and by the step change in performance of our internal model, we launched an effort to evaluate it on all open Millenni
2. 2.▲
   by OneManyNone 2mo ago

   All of that may be true, but pangram currently has a false positive rate of about 1 in 10000, and this has been tested by feeding in thousands of texts written before 2020. That may not last if AI companies start trying to build models that
3. 3.▲
   by OneManyNone 7mo ago

   That’s fair. Clearly Knuth himself thought it was impressive, that’s a strong signal.
4. 4.▲
   by OneManyNone 7mo ago

   Claude did not find a proof, though. It found an algorithm which Knuth then proved was correct.
5. 5.▲
   by OneManyNone 7mo ago

   The companies aren’t changing anything. LLM outputs are just more random than people realize. Run the same prompt 10 times if you really want to know how well they can answer.
6. 6.▲
   by OneManyNone 2y ago

   Counterpoint: What progress has generative linguistics made in the same amount of time that deep learning has been around? It sure doesn't seem to be working well. Also, the racecar example is because of tokenization in LLMs - they don
7. 7.▲
   by OneManyNone 2y ago

   I think this is greatly complicated by the fact that the human brain has been "pre-trained" (in the deep learning sense) by hundreds of millions of years of evolution. A pre-trained LLM also can also learn new concepts from extrem