Emergence of Linguistic Communication from Referential Games with Symbolic and Pixel Input
This research addresses the problem of scaling up emergent communication tasks for AI agents by using deep learning on more realistic pixel inputs, though it is incremental as it extends previous symbolic work.
The study trained reinforcement-learning neural network agents on referential communication games using raw pixel data instead of symbolic inputs, finding that the degree of structure in the input data influences the nature of emerged communication protocols, with structured compositional language more likely when agents perceive a structured world.
The ability of algorithms to evolve or learn (compositional) communication protocols has traditionally been studied in the language evolution literature through the use of emergent communication tasks. Here we scale up this research by using contemporary deep learning methods and by training reinforcement-learning neural network agents on referential communication games. We extend previous work, in which agents were trained in symbolic environments, by developing agents which are able to learn from raw pixel data, a more challenging and realistic input representation. We find that the degree of structure found in the input data affects the nature of the emerged protocols, and thereby corroborate the hypothesis that structured compositional language is most likely to emerge when agents perceive the world as being structured.