I’m less worried about the specific complaints about model deprecation, which can be ‘solved’ for those people by not deprecating the models (obviously costs the AI firms). I’m more worried about AI-induced psychosis.

An analogy I saw recently that I liked: when a cat sees a laser pointer, it is a fun thing to chase. For dogs it is sometimes similar and sometimes it completely breaks the dog’s brain and the dog is never the same again. I feel like AI for us may be more like laser pointers for dogs, and some among us are just not prepared to handle these kinds of AI interactions in a healthy way.