Emergent Multi-Agent Communication in the Deep Learning Era

Abstract

The ability to cooperate through language is a defining feature of humans. As the perceptual, motory and planning capabilities of deep artificial networks increase, researchers are studying whether they also can develop a shared language to interact. From a scientific perspective, understanding the conditions under which language evolves in communities of deep agents and its emergent features can shed light on human language evolution. From an applied perspective, endowing deep networks with the ability to solve problems interactively by communicating with each other and with us should make them more flexible and useful in everyday life. This article surveys representative recent language emergence studies from both of these two angles.

Added some more references and discussion

References in corpus (17)

• Categorical Reparameterization with Gumbel-Softmax
• Learning to Communicate with Deep Multi-Agent Reinforcement Learning
• The Concrete Distribution: A Continuous Relaxation of Discrete Random Variables
• Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments
• Social Influence as Intrinsic Motivation for Multi-Agent Deep Reinforcement Learning
• The Hanabi Challenge: A New Frontier for AI Research
• Cooperating with Machines
• Emergence of Grounded Compositional Language in Multi-Agent Populations
• Multi-Agent Cooperation and the Emergence of (Natural) Language
• On the Utility of Learning about Humans for Human-AI Coordination
• Bayesian Action Decoder for Deep Multi-Agent Reinforcement Learning
• Generative Deep Neural Networks for Dialogue: A Short Review
• Learning to Play Guess Who? and Inventing a Grounded Language as a Consequence
• On the interaction between supervision and self-play in emergent communication
• Biases for Emergent Communication in Multi-agent Reinforcement Learning
• Emergence of Communication in an Interactive World with Consistent Speakers
• Shaping representations through communication: community size effect in artificial learning systems

Cited by in corpus (16)

• Semantics-Native Communication with Contextual Reasoning
• Interpretable agent communication from scratch (with a generic visual processor emerging on the side)
• The Emergence of Wireless MAC Protocols with Multi-Agent Reinforcement Learning
• Truthful AI: Developing and governing AI that does not lie
• Emergent Communication of Generalizations
• Extending rational models of communication from beliefs to actions
• Emergent Discrete Communication in Semantic Spaces
• Grounding Artificial Intelligence in the Origins of Human Behavior
• Quasi-Equivalence Discovery for Zero-Shot Emergent Communication
• Rat big, cat eaten! Ideas for a useful deep-agent protolanguage
• Symbolic Behaviour in Artificial Intelligence
• Low-Bandwidth Communication Emerges Naturally in Multi-Agent Learning Systems
• Reconciling the Discrete-Continuous Divide: Towards a Mathematical Theory of Sparse Communication
• Sparse Communication via Mixed Distributions
• Reinforced Natural Language Interfaces via Entropy Decomposition
• Towards Learning to Speak and Hear Through Multi-Agent Communication over a Continuous Acoustic Channel