12. Risks from Advanced AI

This is the twelfth lecture in the Language Models and Intelligent Agentic Systems course, run by Meridian Cambridge in collaboration with the Cambridge Centre for Data Driven Discovery (C2D3).

This lecture covers risks from advanced AI. These risks might be significant, and if they are, then perhaps AI researchers and broader society ought to care about reducing them. We begin by introducing some useful ways of thinking and reasoning about risk, such as threat modelling and the security mindset. Then, we cover many different threats from AI that have been explored in the literature, like malicious usage, rogue AIs, or gradual disempowerment. Finally, we discuss some of the underlying drivers of AI risk.