Abstract
When AI interacts with humans or other AI systems, new effects emerge. I will discuss three lenses for understanding these effects:
* Misuse: AI unlocks powerful capabilities for good, but the same capabilities can be misused. AI is likely to change the face of cybersecurity, biosecurity, online trust, and other domains.
* Markets: AI systems interact in markets with other AI systems as well as human users; healthy markets produce cheap and performant systems, but unhealthy markets can drive concentration of power or races to the bottom.
* Managers: AI is likely to add significant complexity to society, and we will need a way to manage that. Fortunately, AI systems themselves are well-suited to managing complexity, so we need ways to leverage them to help our own understanding.