Stuart Russell

Stuart Russell is a British computer scientist, professor at the University of California, Berkeley, and co-author with Peter Norvig of Artificial Intelligence: A Modern Approach, the field's standard textbook. In 2016 he founded the Center for Human-Compatible Artificial Intelligence at Berkeley. Human Compatible (2019) argues that the standard way of building AI, giving machines a fixed objective to pursue, becomes dangerous as systems grow more capable. He proposes instead machines that are uncertain about what humans want and learn our preferences from our behaviour, so that they remain open to correction.

Key ideas

• Building AI to optimise a fixed objective is dangerous, because a capable machine pursues a mis-specified goal to harmful extremes.
• A beneficial machine's only objective should be the realisation of human preferences.
• The machine should be initially uncertain about those preferences, which gives it a reason to defer and allow itself to be switched off.
• Human behaviour is the main source of evidence about what humans want.

Major works

• Human Compatible: Artificial Intelligence and the Problem of Control (2019)
• Artificial Intelligence: A Modern Approach (1995)
• Do the Right Thing: Studies in Limited Rationality (1991)

In the Book corner

• Human Compatible: Artificial Intelligence and the Problem of Control 2019 · Beginner
  A book by a leading AI researcher on why ever more capable AI could become dangerous, and how it might be redesigned…

In the conversation

No topics yet. Start one about this thinker.