OPENAI
OpenAI's AI is an insufferable nerd
OpenAI has released a collection of new maths research produced by one of its most advanced internal AI models, with the results now available on GitHub for researchers to review.
The company worked with an independent advisory group at the Institute for Advanced Study to help decide how to share the findings.
Here’s what you should know:
Each result required an average of roughly three hours of ChatGPT Pro thinking power.
OpenAI plans to fund workshops, conferences and research programmes to help mathematicians study AI-generated discoveries.
The company is also working towards releasing the model behind the research, while continuing to test its abilities in maths and other scientific fields.
The square root of showing off
Some of the mathematical proofs are also available in Lean, a programming language that lets computers check whether proofs are correct.
OpenAI is sharing more details about how the model reached its results, including summaries of its reasoning, the number of problems attempted and how much computing power was used.
I bet this model would've cried over getting 99% on a test. - MG