After working with the frontier models for over a year I agree with you. Some problems are "AI shaped" but it's not a universal solution
A single AI will never be a universal solution because there is no universal problem.
Optimistically, and IMO very likely, AI will allow us to continue to create progressively better tools to solve particular problems, however, which is more useful.
Pessimistically, and also IMO very likely, AI will allow us to continue to create progressively bigger problems by allowing us to reach further beyond our grasp, amplifying our abilities without necessarily amplifying our understanding of the consequences of our actions.
On two occasions I have been asked [by members of Parliament], 'Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?' I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question.
It’s been really great for certain problems. It may even “cure cancer.” As it stands, AI is good at specific things and not all types of things. Yet.