Natural Language Does Not Emerge 'Naturally' in Multi-Agent Dialog
This challenges assumptions in multi-agent learning about language emergence, showing it is not automatic and incremental improvements are needed.
The paper tackles the problem of whether natural language emerges in multi-agent dialog without human supervision, finding that most agent-invented languages are effective but not interpretable or compositional, and that human-like language requires increased communication restrictions.
A number of recent works have proposed techniques for end-to-end learning of communication protocols among cooperative multi-agent populations, and have simultaneously found the emergence of grounded human-interpretable language in the protocols developed by the agents, all learned without any human supervision! In this paper, using a Task and Tell reference game between two agents as a testbed, we present a sequence of 'negative' results culminating in a 'positive' one -- showing that while most agent-invented languages are effective (i.e. achieve near-perfect task rewards), they are decidedly not interpretable or compositional. In essence, we find that natural language does not emerge 'naturally', despite the semblance of ease of natural-language-emergence that one may gather from recent literature. We discuss how it is possible to coax the invented languages to become more and more human-like and compositional by increasing restrictions on how two agents may communicate.