AI Risk Assessment
Sayash discusses the challenges of assessing AI risk due to the lack of historical reference classes, making probability estimates speculative. He contrasts AI risks with more established reference classes, such as coin tosses and geopolitical events, highlighting the absurdity of attempting to draw parallels with risks like animal extinction or atmospheric changes. This leads to the conclusion that deductive estimates may be the only viable approach for understanding AI-related risks.In this clip
From this podcast
Machine Learning Street Talk (MLST)
Sayash Kapoor - How seriously should we take AI X-risk? (ICML 1/13)
Related Questions