[Macroagents] 1. The Macroagent Ontology

The blog post on LessWrong introduces the idea of a macroagent, an optimization process made up of agentic subsystems such as teams, companies, markets or AI societies. It breaks each macroagent down into three primitives—memory, mechanisms and the subsystems themselves—illustrating these with examples from the free market, scientific community and government. The author also explains how incentives can be viewed at both the individual subsystem level and the broader role level, and proposes five attributes for comparing macroagents, including composition, size and alignment. In the next installment, LessWrong says the writer will discuss design lenses for optimizing macroagents.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton