I’m reading the published paper by Tomlinson et al now. They don’t know the actual occupations of the users because they anonymized the data. They are inferring the occupations based upon activities requested to AI by 200,000 users and then estimating how many of each occupation is probably in that set of 200,000. So while a lot rings true, it could be completely wrong. However it is interesting to see that much of what historians do are skills AI is doing but also that people aren’t necessarily looking for. The 200k seem more to be asking for information and data, not for analysis. That may change but perhaps people want info and then prefer or expect to think for themselves?