What's the Worst AI Can Do? This Team Is Finding Out.
The Journal.
00:00
Ensuring AI Safety: Benchmarking and Levels
This chapter explores the methods of safety testing for AI models, highlighting the necessity of efficient and straightforward testing processes. It discusses the significance of AI Safety Levels (ASL) in categorizing model risks and raises concerns about self-assessment biases in evaluating AI safety.
Transcript
Play full episode
