Should we go back to Curated Reading Lists?

Werner Kuhn commented on the current overproduction of scientific publications in his Blog and pointed out, correctly, that the new AI technolgoy (e.g. ChatGPT) just regurgitate the texts it was trained on and may only by chance produce new interesting combinations and in general often reasonably sounding uninteresting repetitions of known ideas. He stressed the need to reduced the publication stream, which will be increased with the new means to produce publications fasterA professor, publishing more than 200 original articles a year, was reported to admit that with ChatGT he can produce a paper in a day, when he previously required 2 to 3. Fortunately he was suspended from his university for other misdemeanors …Der Standard

.

Werner stressed the importance of ideas of which current publications often contain very little. Perhaps we should return to the curated reading lists we produced for students decades ago, when access to the literature was difficult; we produced Readers often as reprinted collection of best papers. We hoped the ideas they contained would guide our graduates in their research. I guess the technology today will permit to construct virtual readers, curated by experts (and quickly flooded by AI produced imitations).

A knowledgeable and reputed individual creating a curated reading list points to texts, which contain valuable, new and original ideas. Some of this selection process was done by the review process of journals, when it worked, and even more by scientists writing the text books we learned from. We did it for our students based on our own critical reading.

Such curated reading lists should be seen as suggestions of reading materials others have found worthwile reading. Perhaps help for younger scholars to develop a feel what is important, but certainly not restrict readers to explore on their own and go beyond what the recommendations.

A surprising effect of the web is, that materials published in reputed journals can be found, but for many researchers copies are behind paywalls and therefor not accessible.Or acccess is obtained only after opening a VPN connection, login, etc.

Material in less quality-minded outlets is often easier accessible..

The Guardian reportedon people asking about Guardian articles they were informed about by ChatGT or similar; the articles were unknown to the Guardian, meaning, they were never published, and unknown to the purported authors, i.e. not written by them, but judged by the purported authors that style and content were similar to what they had written. It seems that ChatGPT is capable and willing to invent publications on the fly. Will we have to deal with even more publications than the ones actually published?

The question posed is then, how are we ever sure, that a text is from the author himself and not invented by AI systems? I have control of what is published on my web site and with Pretty-Good-Privacy (PGP)Wikipedia

I can certify that a piece of mail is from me – unless my computer or my account with my internet provider is hacked. A search with ChatGPT however may provide material never written by the purported authors.