ETAMU Librarian Presents “ChatGPT Unveiled” to California Public Library

A.P. Anderson, a research and instruction librarian at East Texas A&M, represented the university on the West Coast this week with a Zoom presentation titled “ChatGPT Unveiled: Understanding the Power, Risks, and Ethics of AI-Language Models.” The presentation was hosted by Livermore Public Library in Livermore, California.

During the one-hour presentation, Anderson discussed the complexities of large language models (LLMs) such as ChatGPT, how theyfunction, and the critical risks and social and ethical dilemmas that have arisen as a result.

Anderson explained that although ChatGPT mimics human interaction, it actually uses predictive math, not logical reasoning, to create responses. As a result, it is prone to oversimplifying complex issues, providing false information, and even fabricating information, known as hallucinating or “confabulating.”

Anderson urged caution when using AI tools like ChatGPT. He suggested some appropriate uses could be asking ChatGPT to brainstorm ideas for a novel or asking it to break down to-do lists into manageable chunks. He warned that higher-stakes tasks, like asking ChatGPT to rank resumes in a job search, assess or diagnose medical conditions, or dispense therapy in place of a qualified human therapist, could all be inappropriate uses and create dangerous outcomes for users.

“These tools are not really up to life-or-death decision-making at this point, but as you're aware of how they work and some of the pitfalls in them, you're going to be able to determine if some use cases have an acceptable level of risk for you,” Anderson said.

Listen to A.P. Anderson's full presentation on Livermore Public Library's YouTube page.