Abstract

Steels & Belpaeme (S&B) refer to the neural plausibility and evolutionary plausibility of their algorithms. Although this is not central to their goal of effective artificial agents, their algorithms are not neurally or evolutionarily plausible. Their communication games are interesting, and more complex games would lead to more effective agents. However, the algorithms could be improved either by using standard subsymbolic algorithms or by algorithms that are really neurally or evolutionarily plausible.