Prioritising new questions over answers is a very good point, which I feel is let down by a somewhat bloated piece dedicated to summarising a series of essays. I think there is more to build on from that premise.
Perhaps the very selling points of "AI" are fallacies. If we turn to chatbots for answers, or to have an image or wall of text composed from a one line prompt, we cheat ourselves of the questions that would arise when we start researching a subject and writing our own notes; or studying an object as reference for a drawing.
But those processes without thought are marketed as time savers, as productivity. That may be an immediate gain — "for whom?" would be a productive follow-up question here — however, it reduces people to button clickers rather than problem solvers.
Filling out a blank space with a lot of words or pixels statistically modeled after the most likely pattern based on a prompt and all the training material in the world — that isn't productivity. It's literally just filling out a blank. I'd compare it to the sort of absent-minded drawing you make while you're on the phone, but phone doodles are more productive!
And if we fill out formulas and spreadsheet unproductively, the result isn't saving time anymore, because we'll need to do it over. And over. And over, until preferably a person will sit down and do the work. Ask the right questions, and then ask follow-ups based on the conclusions.
A hint here, the "right questions" aren't how to improve the model or the prompt, but whether we should be using "AI" for this in the first place.
It's in working through problems that we learn, and jog our wetware minds. It's that endless line of questions that truly generate new knowledge, like a three year old that won't stop asking "why?" to every successive answer you give them. Like the little nag in your brain that keeps working at an issue even though it's 4 AM and you'd really like some sleep.