Does anyone know, or can anyone guess, the business case for predictive text? On phone apps, it is often incredibly difficult to turn off. Why is that, do you think? (The examples I have recent experience with are Facebook and Outlook mobile apps.)
I would have thought that, for AI training purposes, they would want humans typing things and not just regurgitating canned responses. But apparently not?
Might be that information about when you do and don’t use the output is helpful for training. Like, if you use the output, good sign the output is good.