I work in visual AI. I’d like models to be robust enough to be used in practical applications for real, not just founders claiming they can be. Autonomous vehicles, industrial settings, radiology, etc. Current AI (really just machine learning) is very fragile.
It would be good if it could cure cancer, or find missing persons, or improve crop yields or something actually practical. Right now it’s just plagiarizing art and cheating on exams.
Bwahaha, pretty much exactly the conclusion I came to when I got to manage people. People would accept that good advice was valid and useful, then promptly proceed to not follow it. Computers had no opinion on what you told them to do, and proceeded to do whatever you told them to do whether it was good or bad. I decided I’d rather manage computers than people.
My hopes for AI? Jeebus, I have no real idea. Do I want it to behave like a computer or a human? They both suck, in their own way.