Danke, Ich habe Deutsch im Schule gelernt, aber das ist ganz vergessen. I agree, except on the number of jobs :-)

I predict the AI LLM frenzy will stop long before that (both in private and public sector). That won't prevent the slow destruction of universities you document every week, but students/young who can actually read/write/think/count on complex topics will be highly valued.

Super post, but links don't work (you probably put the DOI number but not the URL part).

Totally "fair use", especially as we show how good their products are :-)

Thanks a lot, I will then consider it CC-BY with the bluesky URL - I checked on your website, it is not there.

Fantastic. Is the drawing yours? I would like to use it and correctly cite you.

Fantastic book!

The Choir includes more and more institutions. Once they start taking non-subscription non-APC policies, the effects will become more and more visible

Agreed. Still, we built Matilda on GS model, taking its flaws as things to be fixed (no open data, no access to underlying corpus, untrusted sources, etc.).

Absolutely. I am ok with "recommendation" (as a general term) - even black boxed, as in the end it lets users determine its relevance. This is probably too idealistic, putting the weight on expert users- I don't deal with students.

Ah ah my point was "purely" on basic feasibility, not the actual AI stuff. As you know, that is a territory in which I don't go, I need reproducibility (and transparency) for search/retrieval.

1/ is easy to perform, you search, then you sort/rank 2/ is not trivial at all, adding graph searches is usually quite costly/long.