“LazImpa”: Lazy and Impatient neural agents learn to communicate efficiently

Proceedings of the 24th Conference on Computational Natural Language Learning, pages 335–343 Online, November 19-20, 2020. c⃝2020 Association for Computational Linguistics https://doi.org/10.18653/v1/P17 335 “LazImpa”: Lazy and Impatient neural agents learn to communicate efficiently Mathieu Rita1 Rahma Chaabouni1,2 1Cognitive Machine Learning (ENS/EHESS/PSL Research University/CNRS/INRIA) 2Facebook AI Research mathieu.rita@polytechnique.edu, {rchaabouni,dpx}@fb.com Emmanuel Dupoux1,2 Abstract Previous work has shown that artificial neu- ral agents naturally develop surprisingly non- efficient codes. This is illustrated by the fact that in a referential game involving a speaker and a listener neural networks optimizing accu- rate transmission over a discrete channel, the emergent messages fail to achieve an optimal length. Furthermore, frequent messages tend to be longer than infrequent ones, a pattern con- trary to the Zipf Law of Abbreviation (ZLA) ob- served in all natural languages. Here, we show that near-optimal and ZLA-compatible mes- sages can emerge, but only if both the speaker and the listener are modified. We hence in- troduce a new communication system, “Laz- Impa”, where the speaker is made increasingly lazy, i.e., avoids long messages, and the listener impatient, i.e., seeks to guess the intended con- tent as soon as possible. 1 Introduction Recent emergent-communication studies, renewed by the astonishing success of neural networks, are of- ten motivated by a desire to develop neural network agents eventually able to verbally interact with humans (Havrylov and Titov, 2017; Lazaridou et al., 2017). To facilitate such interaction, neural networks’ emergent language should possess many natural-language-like properties. However, it has been shown that, even if these emergent languages lead to successful communi- cation, they often do not bear core properties of natural language (Kottur et al., 2017; Bouchacourt and Baroni, 2018; Lazaridou et al., 2018; Chaabouni et al., 2020). In this work, we focus on one basic property of nat- ural language that resides on the tendency to use mes- sages that are close to the informational optimum. This is illustrated in the Zipf’s law of Abbreviation (ZLA), an empirical law that states that in natural language, the more frequent a word is, the shorter it tends to be (Zipf, 1949; Teahan et al., 2000; Sigurd et al., 2004; Strauss et al., 2007). Crucially, ZLA is considered to be an efficient property of our language (Gibson et al., 2019). Besides the obvious fact that an efficient code would be easier to process for us, it is also argued to be a core property of natural language, likely to be correlated with other fundamental aspects of human communica- tion, such as regularity and compositionality (Kirby, 2001). Encouraging it might hence lead to emergent lan- guages that are also more likely to develop these other desirable properties. Despite the importance of such property, Chaabouni et al. (2019) showed that standard neural network agents, when trained to play a simple signaling game (Lewis, 1969), develop an inefficient code, which even displays an anti-ZLA pattern. That is, counterintuitively, more frequent inputs are coded with longer messages than less frequent ones. This inefficiency was related to neural networks’ “innate preference” for …

Content truncated to 500 words