I don't have any predictions about when the messy collection of #algorithms we currently call "#AI" will be able to write good #research papers. Could be tomorrow, could be next year, could be a decade, could be never.

But I do know what it will take for me to take the idea seriously: show me an AI that can write a good #literature #review. No hallucinatory references, no "on one hand X, but on the other hand not-X," no repetitive sentence and paragraph structures that make me want to scrub my eyesockets with bleach and a wire brush. It doesn't even have to be publication quality. Just something that would get a decent grade as a term paper in an upper division undergraduate course.

This is the kind of application where AI should be really useful, and right now it's crap. So I'm not holding my breath.