Cars that drive themselves are useful.
Improved inventory tracking is useful.
Improved fraud detection is useful.
Those are a very very minor subset of the vast array of real-world AI use cases.
Edit to respond to your edit :)
Yes ChatGPT is surely overhyped, same as GANs were five years ago, but these things have a whole ton of co-incidencing real-world applications that aren't hyped which we don't see as much. For example facial recognition is so good now that governments are performing mass surveillance with it. Now I admit, this is not in anyone here's opinion a good change. But it is a very real very existing phenomena.
BTW - hope my comment did not come across too confrontational. Just need to defend my AI colleagues ;)
---
Edit to add one more tangental thought that crossed my mind
There's a temptation amongst skeptics to say that technology never lives up to the hype - we were told in the 70s robots would replace us, etc.,
But I think the reason we feel like it never lives up to the hype is because there is a moving target effect taking place. The present will never feel "sci-fi" because as soon as technology comes, the "sci-fi" target moves ahead.
But when you look objectively at the technology around us, there are so many absolutely incredible developments in the last 10-20 years. They just have become normalised.