I have played a little bit with OpenAI's new iteration of #GPT, GPT-o1, which performs an initial reasoning step before running the LLM. It is certainly a more capable tool than previous iterations, though still struggling with the most advanced research mathematical tasks.

Here are some concrete experiments (with a prototype version of the model that I was granted access to). In chatgpt.com/share/2ecd7b73-360… I repeated an experiment from mathstodon.xyz/@tao/1099482491… in which I asked GPT to answer a vaguely worded mathematical query which could be solved by identifying a suitable theorem (Cramer's theorem) from the literature. Previously, GPT was able to mention some relevant concepts but the details were hallucinated nonsense. This time around, Cramer's theorem was identified and a perfectly satisfactory answer was given. (1/3)

I don't know about you but I haven't memorized #WCAG. I got sick and tired of browsing google for answers. I thought that that maybe a good way to go is create a #gpt in the #gptstore. I am also work on a WCAG Reference ap(coming soon)

If you want to try it here is the link.

Feedback is welcome.

chatgpt.com/g/g-MWtM0WK6Z-wcag…

#Wcag2.2

#Accessibility

#A11y

#DigitalAccessibility

ChatGPT - WCAG 2.2 Reference

Your WCAG 2.2 reference for web accessibility guidelines.

^ChatGPT

I have one image, and two very differing descriptions of said image. The first is from #Llava13B, a local model installed on my #Mac, and the second is from #GPT. The GPT response is so huge that I'm going to have to post it as threaded replies, but it shows the vast difference between models. For this first post, the Llava13B response has been included as #AltText.

1/

I know many of you are worried about the potential implications of using #gpt for nefarious purposes as well as its tendency to produce reasonable-sounding bullshit. But I see enormous potential for shitposting. Take this for example, it would have totally averted interplanetary jihad

#Dune #shitposting 7/