OpenAI has released a 35-page paper on Codex (the model that powers GitHub Copilot)! arxiv.org/abs/2107.03374
"We fine-tune GPT models containing up to 12B parameters on code to produce Codex."
They note that GitHub Copilot and the upcoming OpenAI API for the model is powered by descendants of the one in this paper.
They introduce a new dataset of Python programming problems in order to evaluate their models: github.com/openai/human-e…
They report performance by measuring whether or not any one of the k samples pass the unit test for the problem:
They also demonstrate some other interesting applications of Codex like docstring generation. They also provide discussion regarding some of the limitations of the model, like bias, security implications, and legal implications.
Very interesting work!
• • •
Missing some Tweet in this thread? You can try to
force a refresh
I find it very interesting that Twitter recommends relevant tweets to me, but the topic suggestion is completely off. It looks to me like the recommendation and topic selection algorithm are completely different.
While the tweet recommendation algo is more sophisticated that likely takes into consideration the semantic content of the tweet, the topic selection algo seems to be a simple algorithm that heavily weighs the presence of keywords.
Saw few tweets on pigeon-based classification of breast cancer (@tunguz@hardmaru, @Dominic1King, & ML Reddit), which was published in 2015. I work with the legend himself @rml52! I thought for my 1st Twitter thread I'd go over the papers's main points & our current work! (1/11)
My PI often likes to say AI stands for avian intelligence. And indeed his paper shows pigeons can learn the difficult task of classifying the presence of breast cancer in histopathological images. (2/11)
The pigeons were placed in an apparatus and the 🔬 image was shown to the pigeons on a touchscreen. The pigeons were given food if they pressed the correct button on the screen. (This is opposed to regular pathologists who are not given free food when analyzing images!) (3/11)