This isn't true though.
Yes, there are too many products being build that don't serve anyone's needs or solve anyone's problems.
However, many of the AI products do solve problems and serve needs.
You're right though, to compare this to other booms, which also had the same problem. This is very much a "hyper" version, which is pretty incredible to be in the middle of.
If you have a product that:
1. Solves a real problem people would pay for
2. Is not trivially replicable by your potential customers or competitors
3. Does not have a natural discovery mechanism by potential customers
Then you need the marketing budget.
That is not most people's problem.
Unless your thing has strong network effects or a large capex requirement (ex: GPU infra) its easily replicated and I think that's really what makes things hard.
It will take a few years for investors to figure this out, but in the meantime, everyone is spreading their bets around like peanut butter in order to be in the game.
Every solution to a problem comes with its own costs. It is entirely possible that most solutions that are rooted in modern computing technology have actual or perceived costs that exceed the value of "solving the problem".
The problems that most people have that they really want to solve are not addressable by AI, or computers, or software.