
0:00
28:13
If you want an AI to "play nice", you need to give it values to ensure it is acting beneficially toward humans. To solve this alignment problem we need to know what values we are talking about. Today I describe a five-level classification system which helps us to understand all values and how they relate to one another. These include survival, cognition, those related to individuality and sociality, and a concern for the wider environment. We then look at how to apply these to AI systems and examine several critiques of the theory.
More episodes from "Save the World with Jay Friedenberg"



Don't miss an episode of “Save the World with Jay Friedenberg” and subscribe to it in the GetPodcast app.








