Which mentor specializes in scalable oversight in AI safety?

• A. Sam Bowman
• B. Peter McCrory
• C. Saffron Huang
• D. Nicholas Carlini

Answer: (A)

Scalable oversight in AI safety is about creating ways to supervise and steer very capable AI systems even when a single human can’t monitor every action. It asks how we design tasks, feedback, and evaluation so oversight remains feasible as models scale—often using decomposition of problems, iterative feedback loops, crowdsourced or simulated evaluation, and techniques like reward modeling that stay reliable when the model’s outputs become more complex. The mentor who specializes in this area is Sam Bowman, because his work focuses on building and evaluating supervision and alignment processes that can operate at scale, helping researchers develop effective oversight mechanisms without requiring proportional human labor. The other mentors are known for different topics within AI safety and security, rather than scalable oversight, so they don’t fit this specialization as directly.