Enhancing AI Decision-Making with Wise Values

17min Snip

00:00

Play full episode

Summary

Transcript

Episode notes

This chapter delves into training AI models to act on values that align with wisdom rather than just losses, using techniques like RLHF and Constitutional AI. It discusses the importance of understanding and representing agreed-upon values to improve decision-making processes in various contexts, including in politics and social media. The chapter also explores the application of technology to facilitate constructive collaborations among individuals with differing viewpoints, emphasizing the significance of personal values in shaping discussions and decision-making.

Read the full transcript here.

What are the best ways to define "values" and "meaning"? How can democratic processes harness people's intrinsic values and sources of meaning to increase their agency, cooperation, participation, equality, etc.? To what extent do political rivals — or even the bitterest of political enemies — actually value many of the same things? Might we be able to use AIs as "neutral" third-party mediators to help reduce political polarization, especially on an interpersonal level? How can we transform our personal values and sources of meaning into positive, shared visions for society? Are markets inherently antisocial? Or are they just easily made to be so? Companies frequently invoke our deepest needs and values as a bait-and-switch to sell us their goods and services; but since there must actually be demand to have those deep needs met and those deep values realized, why do companies so rarely attempt to supply goods and services that address those things directly? Assuming there actually is a lot of overlap in intrinsic values and sources of meaning across individuals and across groups, why do we still have such a hard time developing shared visions for society?

Joe Edelman is a philosopher, sociologist, and entrepreneur. He invented the meaning-based metrics used at CouchSurfing, Facebook, and Apple, and co-founded the Center for Humane Technology and the Meaning Alignment Institute. His biggest contribution is a definition of "human values" that's precise enough to create product metrics, aligned ML models, and values-based democratic structures. Follow him on Twitter / X at @edelwax, email him at hello@meaningalignment.org, or learn more about him and the Meaning Alignment Institute (@meaningaligned on Twitter / X).

Further reading: