Peter Leonard
Associate University Librarian for Research Data Services
Stanford University
Although commercial large language models such as ChatGPT have captured the public imagination, the math underlying these conversational systems has important implications for libraries in key areas such as speech-to-text, image description, and handwriting recognition. Collections previously isolated from text-mining practices now stand at the cusp of becoming computationally tractable. Also, in an era of increased attention to accessibility requirements, providing textual proxies of audio-visual and image-based content can lessen legal exposure. The session will consider applications of transformers-based artificial intelligence models in an international perspective, with a focus on what they enable and what changes they imply for digital libraries.