Comment by kleiba2

1 day ago

Interesting that AI apparently doesn't write papers for humans: even experts struggle to understand them. But other AIs do. So AI basically writes papers for other AIs.

That's what reinforcement learning does when done this way, the AI is literally trained to write to itself in the way that will aid it to get to a solution most efficiently. I assume they still have some additional fine-tuning on top to keep the writing somewhat natural, but still.