OpenAI math
As AI technology advances, we're getting closer and closer to 'solving' one of the hardest fields humans have ever touched: Math
OpenAI just published 772 manuscripts organised into 372 families of math proofs
Language models seem to be really good at solving problems that demand logical reasoning, and this is obvious when you understand that reinforcement learning (learning through rewards) is being done directly in math and logic.
How can they not be better than us? More than 100,000 NVIDIA GPUs working together to bring to 'life' this reasoning machine with one sole objective of becoming better at reasoning tasks and text generation. And, as they become smarter, the harder it gets to understand their proofs, but this OpenAI tweet shows us where we're heading.
I only created this post because I was surprised. Math changed so much recently, and mathematicians don't know how to deal with their field changing.
The same happened to art, software, and text, and it seems like it will continue happening to other fields.