Personally, I think that the true existential risk posed by AI, is that the systems will be shitty, not that they will become so super-advanced whatever whatever. But that they will just sort of chug along, always half-broken, always/never being repaired or improved, always giving you a crappy automated alternative of something you used to be able to get by communicating with a real person. To me, that’s much more terrifying than any kind of Skynet scenario.