Learning to Reason with LLMs

Post by aantix » Thu, Sep 12, 2024, 5:36 PM UTC

apsec112 wrote: post #71

rfw300 wrote: post #39

A lot of skepticism here, but these are astonishing results! People should realize we’re reaching the point where LLMs are surpassing humans in any task limited in scope enough to be a “benchmark”. And as anyone who’s spent time using Claude 3.5 Sonnet / GPT-4o can attest, these things really are useful and smart! (And, if these results hold up, O1 is much, much smarter.) This is a nerve-wracking time to be a knowled…

Even without AI, it's gotten ~10,000 times easier to write software than in the 1950s (eg. imagine trying to write PyTorch code by hand in IBM 650 assembly), but the demand for software engineering has only increased, because demand increases even faster than supply does. Jevons paradox: https://en.wikipedia.org/wiki/Jevons_paradox

Post by seydor » Thu, Sep 12, 2024, 5:37 PM UTC

valine wrote: post #8

The model performance is driven by chain of thought, but they will not be providing chain of thought responses to the user for various reasons including competitive advantage. After the release of GPT4 it became very common to fine-tune non-OpenAI models on GPT4 output. I’d say OpenAI is rightly concerned that fine-tuning on chain of thought responses from this model would allow for quicker reproduction of their resu…

Post by rfw300 » Thu, Sep 12, 2024, 5:37 PM UTC

awesomeMilou wrote: post #57

rfw300 wrote: post #39

A lot of skepticism here, but these are astonishing results! People should realize we’re reaching the point where LLMs are surpassing humans in any task limited in scope enough to be a “benchmark”. And as anyone who’s spent time using Claude 3.5 Sonnet / GPT-4o can attest, these things really are useful and smart! (And, if these results hold up, O1 is much, much smarter.) This is a nerve-wracking time to be a knowled…

Is it? They talk about 10k attempts to reach gold medal status in the mathematics olympiad, but zero shot performance doesn't even place it in the upper 50th percentile. Maybe I'm confused but 10k attempts on the same problem set would make anyone an expert in that topic? It's also weird that zero shot performance is so bad, but over a lot of attempts it seems to get correct answers? Or is it learning from previous a…

Post by rankam » Thu, Sep 12, 2024, 5:38 PM UTC

p1esk wrote: post #58

Do people see the new models in the web interface? Mine still shows the old models (I'm a paid subscriber).

Post by TheAceOfHearts » Thu, Sep 12, 2024, 5:38 PM UTC

zozbot234 wrote: post #49

TheAceOfHearts wrote: post #26

Kinda disappointed that they're hiding the thought process. Hopefully the open source community will figure out how to effectively match and replicate what OpenAI is doing. I wonder how far we are from having a model that can correctly solve a word soup search problem directly from just a prompt and input image. It seems like the crossword example is close. For a word search it would require turning the image into an…

> Hopefully the open source community will figure out how to effectively match and replicate what OpenAI is doing. No need for that, there is a Reflection 70B model that does the exact same thing - with chains of thought being separated from the "final answer" via custom 'tag' tokens.