Societies of Intelligent Agents

Abstract

Abstract.A definition of language is proposed in which language is a lowbandwidth channel that can increase agent rewards in a reinforcement learning setting, and in which agents can learn to produce language and teach it to other agents. Societies of agents are being modeled by economists to understand economic instability and other non-equilibrium phenomena. I hypothesize a divergent distribution of intelligence in societies of agents when rewards can be exchanged for increases in agent information processing capacity.

AGI and the Knight-Darwin Law: why idealized AGI reproduction requires collaboration

“…More practically, Theorem 9 might suggest prioritizing research on multiagent approaches to AGI, such as [6], [12], [14], [17], [19], [21], and similar work.…”

MAS-based Agent Societies by Means of Scout Movement

“…The needed tools are the role, the mode, and the strategy of cooperation. The coherence can be measured by three properties namely the completeness, efficiency (a local goal will be realized by only one agent), and consistency (no agent of the society can destroy the results achieved by another agent) of goals (Hibbard, 2011).…”

“…On the other hand, a divergent distribution of intelligence in societies of agents where rewards can be exchanged for increases in agent information processing capacity is hypothesized in Hibbard (2011). This paper poses questions about societies of intelligent agents using a model based on Reinforcement Learning (RL) (Hutter, 2004;Legg & Hutter, 2006).…”