Most non-trivial expertise topics are not one-dimensional. You might be at the "ceiling" in some particular sub-niche, while still on the floor on other aspects of the topic.

So even if you follow the artcles premise (I do not), it still can potentially 'raise' you wherever you were.

Key seems to be wether you have enough expertise to evaluate or test the outputs. Some others have refered to this as having a good sense of the "known/unknown" matrix for the domain.

The AI will be most helpful for you in the known-unknown / unknown-known axis, not so much in the known-known / unknown-unknown parts. The latter unfortunatly is were you see the most derailed use of the tech.