I use AI tools in my daily work and, for sure, it can boost my productivity as much as 10x. For me, this is a very high level of productivity because my baseline was already high; call it "10", boosted to "100". Being a natural 10 means that I know how to construct good prompts, give the LLM hints and examples, etc., that someone who is a 3 wouldn't know.
Here's the problem, though. If you rely on AI for a "shortcut" all the time, your baseline barely increases because you're not teaching your real brain the fundamentals of your technical domain, be it accounting or computer programming or medicine or whatever. Someone who is a 10/100 over a year might get their productivity to an 11 /110 by learning through use of the LLM (or it's possible the trend is you de-learn and then your productivity goes to 9 /90). But, someone who is a 3 could spend that same time intensely studying the fundamentals to get their natural level to a 12, making them a 12/120.
This is why the idea that AI tools could qualify anyone to perform any job, e.g. a random guy off the street to be a rocket scientist, is nonsense!
I feel that it's very dangerous for the economy if we're trying to replace too much actual thinking with AI. The productivity boost will be short-term and then could drop off a cliff. My fear isn't that AI will replace jobs, but that
The objection would be that the AI boost will increase faster than we will get dumber. For example, someone who is a 10/100 and would dumb down to an 8/80 if the boost stayed at 10x will actually stay 100+ because the next AI model will give him a 13x boost to 8/104. I think it's obvious why this won't work long-term.