> doesn't mean that you can actually deliver on that promise

It's very rare that the big risk in a problem is technical feasibility. And you should definitely be, or partner with, someone who can tell the difference between a spec that can be built to a spec that has real technical risk, and only sell things that you can deliver.

> The problem is that you don't know what people want. People don't know what they want

Indeed, and the only real way to tell is to build it. But you can de-risk by selling a few first and then building, instead of just investing years of your life completely blindly. And there's no reason not to. It's very rare that a product idea can be sold once it's built but not before, and there are other, less risky, products you could build instead of that one.