Generative AI 101

Generative AI is many things: it is simultaneously scary, useful, flexible, opaque, exciting, and dangerous. Researchers, policymakers, creatives and executives around the world are turning to these models to answer complex questions, rethink practices and procedures, and save time. Large language models are performing scientific research in many fields, while democratizing access and understanding of some types of AI. Yet researchers struggle to explain what these systems are actually doing, and how they utilize personal, public and proprietary data to answer our prompts. The data giants with huge computing power, skilled workers, and large troves of data are the main suppliers of generative AI and reaping many of the potential income and investments. Moreover, these systems are not perfect: they  make mistakes, do not have real time information and can perpetuate inaccuracies and disinformation. Finally, there is growing evidence that these chat-bots can both upskill (help less educated workers be productive) and deskill. In this webinar, we asked three researchers of varied backgrounds (computer science, communications, and systems engineering)  to discuss generative AI–its potential and pitfalls. We engaged in a moderated discussion for about 30 minutes and then we opened up the floor to audience questions.

Speakers:

• Hal Daumé III, Ph.D. Volpi-Cupal Family Endowed Professor in Computer Science, University of Maryland, and co-author of Language (Technology) is Power
• David Broniatowski, Ph.D, Associate Professor, Engineering Management and Systems Engineering, Director of the Decision Making and Systems Architecture Laboratory at GWU,and author of  Psychological Foundations of Explainability and Interpretability in Artificial Intelligence 
• Sarah Myers West, Ph.D., Managing Director of AI Now Institute, and co-author of Confronting Tech Power

Moderator:

Dr. Susan Ariel Aaronson, Director, Digital Trade and Data Governance Hub, Research Professor and Cross-Disciplinary Fellow, GWU