Judging by the abstract in that linked paper, I think that it is arguing more about semantics than anything else. It basically claims that: "No person is truly self taught as they stand on the shoulders of their predecessors". To which, I would agree. No man is an island and it would be very difficult to teach oneself something without material composed by others. However, I think the colloqual use of the term autodidact is more about the lack of a formal educational system or formal teaching figure and less about a person figuring everything out on their own.

No one has actually provided a counter argument. A list of 1000 over billions of humans just shows how rare this idea is. That article isn't arguing semantics, and as a matter of a fact it is showing that the truthiness of an autodidact is to be called into question given the nature of mastering a domain.