Journal, week 10
I began this week sitting in on a joint meeting with the cataloging staff of the John Carter Brown Library and the John D. Rockefeller Jr. Library at Brown ("the Rock," as my colleagues here call it). It was meant to be a meeting to discuss a new digital collection platform that the two libraries might consider adopting.
This new platform ended up being Alma Specto, a digital management software developed by Ex Libris, the company that creates the Alma software that JCB already uses for its collection management. A key difference here is that Alma Specto uses an AI large language model to automatically generate metadata for new acquisitions.
As soon as the presenter mentioned that Alma Specto offered AI-based tools for catalogers, several of the participants in the Zoom meeting spoke up to offer significant concerns. One of the participants, who had been present when Ex Libris offered the libraries a software demo, mentioned that he noticed several inconsistencies and errors in the software's generated description. I can remember him saying that the software couldn't differentiate between text that was printed as content of the item and a stamp that had been placed next to the body---the AI-generated transcription simply printed all of the text together. For a rare books library, this is really unacceptable, as even small inaccuracies can make a transcription useless...
I was heartened to hear this very vocal feedback from several librarians because I had been very surprised to hear the presenter even offering the AI tools as a serious consideration for the libraries. But, of course, AI is winding its way into every industry, the directors at my current job are even considering ways to integrate it into their database software.
Otherwise, my workload today was pretty typical. I worked on cleaning up records for cataloger's reference items, and got to work with more maps. I am really interested in how complex and full of historical context they are.