But do you ever get good enough to make a contribution that's worthwhile? And also is your knowledge flawed because the knowledge of the data used to train the model is flawed?
Whilst I do see your point and I do see the value for prototyping, I don't quite agree that you can learn very much from it. Not more than the many basic intro to..... Articles can teach.
I can ask the LLM infinite "stupid questions".
For all the things I know a little about, it can push me in the direction of an average in that field.
I can do lots of little prototypes and find the gaps then think and come back or ask more, in turn I learn.