It's easy to imagine how AI can be beneficial in the short term. The problem is imagining how it won't go wrong in the long term.
Even sci-fi has a hard time figuring that out. StarTrek just stops at ChatGPT-level of intelligence, that's how smart the ship computer is and it doesn't get any smarter. Whenever there is something smarter, it's always a unique one-of that can't be replicated.
Nobody knows how the world will look like when we have ubiquitous smart and cheap AI, not just ChatGPT-smart, but "smarter than the smartest human"-smart, and by a large margin. There is basically no realistic scenario where we won't end up with AI that will be far superior to us.