Towards multi-agent emergent communication as a building block of human-centric AI

Angeliki Lazaridou / DeepMind

Apr 26, 2021

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 can also develop a shared language to interact. In this talk, I will highlight recent advances in this field but also common headaches (or perhaps limitations) with respect to experimental setup and evaluation of emergent communication. Towards making multi-agent communication a building block of human-centric AI, and by drawing from my own recent work, I will discuss approaches on making emergent communication relevant for human-agent communication in natural language.

Bio: Angeliki Lazaridou is a research scientist at DeepMind. Before that, she was a graduate student of Marco Baroni working on grounded language learning at the CLIC Lab of the Center for Mind/Brain Sciences of the University of Trento, Italy. Before that, she did an MSc in Computational Linguistics at the University of Saarland working with Ivan Titov and Caroline Sporleder on Sentiment Analysis, supported by an Erasmus Mundus Masters scholarship in Language and Communication Technology (EM-LCT).