At the 2020 meeting of the American Political Science Association (APSA), the largest annual meeting of political scientists in the US, Stuart Russell, professor of computer science at UC Berkeley and co-author of the standard textbook on AI, “Artificial Intelligence: A Modern Approach”, discussed his recent book, “Human Compatible” , with three critics:
Allan Dafoe, Director of the Centre for the Governance of AI at FHI, and Associate Professor in the International Politics of AI at Oxford University
Rob Reich, Director of the McCoy Family Center for Ethics in Society, and Professor of Political Science at Stanford University
Marietje Schaake, International Policy Director at the Cyber Policy Center at Stanford University, and former Member of the European Parliament
The YouTube description provides time stamps for important transitions within the video. During the roundtable, Russell argues that to make AI compatible with human existence, we must abandon the notion of AIs acting on fixed objective functions in favor of AIs (having uncertainty about their objective functions) consulting humans to update their objective functions. The critics then provide their responses to Russell’s thesis, which is followed by a general Q&A.
Stuart Russell Human Compatible AI Roundtable with Allan Dafoe, Rob Reich, & Marietje Schaake
At the 2020 meeting of the American Political Science Association (APSA), the largest annual meeting of political scientists in the US, Stuart Russell, professor of computer science at UC Berkeley and co-author of the standard textbook on AI, “Artificial Intelligence: A Modern Approach”, discussed his recent book, “Human Compatible” , with three critics:
Allan Dafoe, Director of the Centre for the Governance of AI at FHI, and Associate Professor in the International Politics of AI at Oxford University
Rob Reich, Director of the McCoy Family Center for Ethics in Society, and Professor of Political Science at Stanford University
Marietje Schaake, International Policy Director at the Cyber Policy Center at Stanford University, and former Member of the European Parliament
The YouTube description provides time stamps for important transitions within the video. During the roundtable, Russell argues that to make AI compatible with human existence, we must abandon the notion of AIs acting on fixed objective functions in favor of AIs (having uncertainty about their objective functions) consulting humans to update their objective functions. The critics then provide their responses to Russell’s thesis, which is followed by a general Q&A.