How AI Fails Us, and How Economics Can Help

On Wednesday, April 19, Michael I. Jordan, Pehong Chen Distinguished Professor in the Department of Electrical Engineering and Computer Science and in the Department of Statistics at UC Berkeley, delivered the third of four Distinguished Lectures on the Status and Future of AI, co-hosted by CITRIS Research Exchange and the Berkeley Artificial Intelligence Research Group (BAIR).

Combining the field of economics — which tends to focus on the collective — with statistics and computer science, Jordan’s lecture aims to provide a broader conceptual foundation for emerging real-world artificial intelligence (AI) systems.

He characterizes AI as not the discovery or creation of a new spark of intelligence — that hard-to-define quality that includes the ability to apply knowledge and elements of reasoning — as it is so often depicted in popular culture. Rather, he sees AI as reflecting the emergence of a new field of engineering that links humans through large-scale systems by combining social science, economics and machine learning. 

Jordan argues that the current state of AI overrates autonomy, with a focus on intelligence existing in a single decision-making entity, or agent. This results in a shallow approach to social situations. Instead, he said, agents should be interactive; intelligence is just as crucial to the collective as it is to individual agents.

“Intelligence—we don’t really know what it is, but it doesn’t just have to do with the human brain or mind. Markets are intelligent … in ways that we aren’t individually,” Jordan said. 

“We study the ant colonies, and we think about the markets,” he said, “but we don’t realize that we could create brand-new kinds of collectives that could be really amazing, instead of trying to replace a single human being with this computer. Our goals and our aspirations should be at the level of collectivity.”