title: ai4lam Metadata/Discovery WG Monthly Meeting

Apr 9, 2024

8 AM California | 11 AM Washington DC | 4 PM UK | 5 PM Oslo & Paris

Attending

• Jeremy Nelson (Stanford)

• Andrew Elliot

• Sara Amato

• Joy Panigabutra-Roberts (University of Tennessee)

• Sarah Mann

• Erik Radio (University of Colorado)

• Ian Bogus (ReCAP)

• Craig Rosenbeck

Helpful Links

• Metadata WG Zotero Group Library

Project Documents and Data

• WG charter

• WG Google Drive folder

Agenda

• Announcements

  • Next meeting will be on May 7 2024 for Abigail Potter presentation on AI at LOC

• Joy Panigabutra-Roberts’ presentation on AI authors and performers in the context of identity management.

  • Head of Cataloging at the University of Tennessee Libraries

  • What about AI Authors and A Robot Comedian?

    • Beta Writer and Steffen Pauly, Lithium-Ion Batteries: A Machine-Generated Summary of Current Research, 2022. From Artificial Intelligence in Libraries and Publishing

    • Brent Katz, Josh Morgenthua, and Simon Rich, I Am Code: An Artifical Intelligence Speaks. Poems by code-devinci-002. 2023. Simon Rich. New York, NY: Back Bay Books.

    • Jon the Robot (comedian)

  • PCC does not consider AI to be authors

  • Consider a Named AI or generative computer program used to create a resource to be a related work, not as an agent…

  • AstroLLLaMA-Chat - https://huggingface.co/universeTBD

    • The first open-source conversational AI tool tailored for teh astronomy community https://doi.org/10.48550/arXiv.2401.01916

  • Excerpt from Youtube (https://www.youtube.com/watch?v=OkCoTixo-MM - A Bot and Costello - Let’s Power the Whole Thing off

  • Philosophical and Legal implications are greater than cataloging these works generated by AI

  • Questions:

    • Amazon generated books?

      • Case in wikipedia articles - predatory publishiser, declined to cataloged to. WIkipedia articles have varied quality,

      • Comes down to collection development to screen out poor-quality

      • Attributions, but include disclaimer, assisted by AI generated,

      • Publisher, authors not claim attribution if generated by AI.

      • How do you screen for these works? Be up front about how you are using generated

    • Not too hard to tell now if work is AI generated, aesthetical judgements by Cataloger. How would cataloger training be in 5 years?

      • Poor quality, don’t catalog

      • Responsibility of Author and Publisher to disclose and up front they need transparency.

    • Run into this issue in your institutions?

      • OCLC RLP Metadata Managers Focus Group will discuss on a topic related to AI and cataloging/metadata with both domestic and international institutions next week

      • Creating cataloging records on ebooks, convert publisher data, Ex Libris training a model using PDFs to train to extract publisher information from records

      • Another presentation Joy attended recently on a library that did an experiment to use ChatGPT with OCLC records, once you feed the data to the model, the data is now in the training set for OpenAI