Thread by @AndrewLampinen

What are symbols? Where do symbols come from? What behaviors demonstrate the ability to engage with symbols? How do the answers to these questions impact AI research? We argue for a new perspective on these issues our preprint: arxiv.org/abs/2102.03406 Summary in thread: 1/6

We interpret symbols as entities whose meaning is established by convention. We therefore argue that an entity is a symbol *only* to a system that demonstrates active participation in a system of meaning by convention; that is, a system that exhibits symbolic behavior. 2/6

What is symbolic behavior, and how could an AI (or animal) demonstrate it? We outline a set of criteria based on our interpretation of symbols: symbolic behavior is receptive, constructive, embedded, malleable, separable, meaningful, and graded. We illustrate with examples. 3/6

How can this symbolic behavior come about? We suggest that social and cultural forces encourage this behavior in humans, and so AI researchers should focus on these areas to encourage the emergence of symbolic behavior. We suggest that this will allow for more capable AI. 4/6

We also discuss the relationship between our perspective and other views. In particular, we argue that the more restrictive definitions of symbols underlying GOFAI and some neuro-symbolic models are limiting. The field should focus on social forces rather than mechanisms. 5/6

Thanks to my co-authors @santoroAI, @korymath, Tim Lillicrap, and David Raposo, and our wonderful colleagues who read earlier drafts of this paper! 6/6

Our argument draws on fields ranging from cognitive science to semiotics to developmental psychology to linguistics, so I'm excited to hear others' thoughts! (E.g. @hawkrobe @natvelali, ...)