I think this assumption is not true anymore - ChatGPT can give an answer in less time and (human) effort than it takes to formulate and devise a question. So it really is about creating a set of curated question-answer pairs, where both the questions and the answers can be auto-generated. In this context, asking the same question different ways is really helpful to an AI. And there is no such thing as a bad question, there are just Q-A pairs that do not contribute significantly to training the model ("boring" questions). I've noticed Quora is experimenting with ChatGPT but they aren't at the point where you can flag the AI's answer as incorrect or misleading and fix it.

No, asking bad or thoughtless questions is easy. Asking a good question, complete with "this is what I've tried" and snippets and links to all revelant resources like API documentation, CodePens/Godbolts/standards and a summary of your thinking that has led to needing an answer in the first place is very hard.

A good question is probably at least an hour's work, and there are very few easy questions left that don't need substantial work even to frame.

An answer can be very in-depth with research and analysis, which also takes a long time, but it can also be anywhere on the spectrum down to fairly straightforward "you missed a function call" answer.

On the other hand, writing a good question sometimes solves your problem half way through when your Minimal Viable Example suddenly starts working and your realise the flaw, or the rubber-duck nature makes you think it through in a new way.