Pandora's box or cornucopia - should we accelerate open access to feed Large Language Models?

• Main Author: TAY, Aaron
• Format: text
• Published: Institutional Knowledge at Singapore Management University 2023
• Subjects: Artificial Intelligence and Robotics; Library and Information Science; Scholarly Communication
• Online Access: https://ink.library.smu.edu.sg/rstf2023/program/agenda/7; https://ink.library.smu.edu.sg/context/rstf2023/article/1003/type/native/viewcontent/4_Aaron.mp4; https://ink.library.smu.edu.sg/context/rstf2023/article/1003/filename/0/type/additional/viewcontent/4_AaronTay_RSTF_AI.pdf
• Institution: Singapore Management University

The latest generation of Large Language Models (LLM) like ChatGPT seem to be the real deal, demonstrating capabilities that show huge leaps in Natural Language Understanding and Generation from past generations of systems. But like all machine learning applications, training on large amount of data is necessary for success, and it seems obvious that this makes open access even more valuable now. Open Access potentially unlocks value not just from human reading it but also machine ones. But should librarians really push and pay more for Open Access? In this short 25 minute talk, I will cover some arguments both for and against this view.