Learning to model other minds
OpenAI Blog
An algorithm called Learning with Opponent-Learning Awareness (LOLA) has been developed to enable agents to account for the fact that other agents are simultaneously learning. LOLA discovered self-interested yet collaborative strategies such as tit-for-tat in the iterated prisoner's dilemma without being explicitly programmed to do so. The approach represents progress toward building artificial agents capable of modeling the mental states and learning processes of other agents.
Why it matters
We’re releasing an algorithm which accounts for the fact that other agents are learning too, and discovers self-interested yet collaborative strategies like tit-for-tat in the iterated prisoner’s dilemma. This algorithm, Learning with Opponent-Learning Awareness (LOLA), is a small step towards agents that model other minds.
