The self driving car industry has been working hard to develop practical applications for AI as it exists in it's current state. Waymo is in the lead, and has been at it since 2009. They are making it happen, but it's hard, the going is slow.

Roboticists have spent decades trying to teach robots to stack boxes, and again, still not ready for primetime. Same with conversational AI. That's just the nature of it. It blows your mind one moment and wets the bed the next.

If Open AI or it's competitors had it in good confidence there was a way to get AI to perform consistently in a single marketable application, they'd be all over it, but that's just not how it works. We can watch China's robots perform all kind of cool tricks, but they still aren't useful for much of anything except entertainment.