The only thing predictable here is AI devs will start having existential meltdowns trying to make things predictable and controllable. language + emotion + social context = chaos.

Engineers historically are the wrong folk to handle such things (chaos management). Expect more Philosophers/Psychologists to start getting roped in. And lot of failed experiments treating people like lab rats.