#72 - Miles Brundage and Tim Hwang

Fairness

Ensuring fairness and accountability in AI systems is a multifaceted challenge. Miles Brundage emphasizes the importance of transparency and cooperation to avoid international conflicts, drawing parallels to historical arms control agreements 1. He suggests that developing governance methods for current AI issues can set positive precedents for future challenges 1. Tim Hwang adds that many fairness problems are essentially value alignment issues, where systems fail to behave consistently with human values 2.

If you actually had the full development of the fat methods and you had accountability and transparency for even general AI systems or super intelligent systems, I think that would open up the door for a lot more collaboration.

--- Miles Brundage

This approach could foster trust and collaboration in AI development, potentially mitigating risks and enhancing benefits.

   

Value Alignment

Value alignment in AI systems is crucial for long-term safety. Miles Brundage discusses the concept of corrigibility, where AI systems are designed to take critical feedback and adjust their actions accordingly 3. This could make accountability more manageable, even for powerful systems. He also notes that solving near-term issues can build a foundation for addressing long-term challenges 1.

If a system is designed in such a way that it's able to take critical feedback and it's able to say, okay, yeah, what I was doing was wrong, that might stabilize in a way that it's continuously asking for humans feedback.

--- Miles Brundage

Tim Hwang highlights the importance of establishing norms within the research community to ensure the development of corrigible systems, which could influence the design of future AI technologies.