Hacker News Reader

I wonder what would happen if we used RL to minimize the user's cognitive debt. Could this lead to the creation of an effective tutor model?

Definitely, but it won't be a creation of any known AI companies, anytime soon. I have a hard time to see how this would be profitable.

It also goes against the main ethos of the AI sect to "stress-test" the AI against everything and everyone, so there's that.