INTERDISCIPLINARY CENTER FOR BIOETHICS: TECHNOLOGY & ETHICS WORKING GROUP
Abstract: ChatGPT and other large language models have taken the world of AI by storm. These systems hold promise for a wide range of applications, but adopting them also carries important risks. On the positive side, large language models can perform impressive feats, such as composing acrostic poems or generating runnable computer code. On the negative side, they also display some surprising shortcomings, such as an inability to reliably count how many items are in a list. How can we explain the seemingly contradictory nature of these systems? In this talk, Dr. McCoy will build an intuition for how large language models work by discussing the four main ingredients that contribute to their success: neural networks, next-word prediction, big data, and human feedback. Understanding these components will provide a lens that helps to explain both the impressive abilities and the surprising failures of these systems - an understanding that is important for recognizing what they are useful for and what they are not.
Please email Working Group Chair Joe Carvalko to receive an event announcement with the link to join on Zoom.
The Technology and Ethics study group examines crucial societal, ethical, and public policy questions arising from the adoption of new technologies. To subscribe to the mailing list for this group, please visit https://mailman.yale.edu/mailman/listinfo/techethics.