Thread by @SamBolotta

Social Neuro AI: Social Interaction as the "dark matter" of AI

Happy to share this opinion paper I wrote with @introspection, on how we can advance the field of #SocialNeuroAI: arxiv.org/abs/2112.15459

A thread (1/10)

Learning adaptive information from others results in better regulation of task performance, especially by gaining fitness benefits and in avoiding some of the costs associated with asocial, trial-and-error learning.

(2/10)

Various categories of #sociallearning have been proposed, as well as social learning strategies that refine such categories and make them contextually appropriate.

(3/10)

Moreover, recent efforts in #computationalsocialneuroscience have paved the way for a naturalization of social interactions, showing that the connectome seems to facilitate resonance between brains.

(4/10)

So, how can we leverage social learning to make AI agents more robust and flexible? Three ideas:

(5/10)

1 Neuroscientific theories of cognitive architecture can enhance #biologicalplausibility and help us understand how we could bridge individual and social theories of intelligence (6/10)

2 Intelligence occurs in time as opposed to over time, and this is naturally incorporated by the powerful framework offered by #dynamics.

(7/10)

3 #Socialembodiment has been demonstrated to provide social interactions between virtual agents and humans with a more sophisticated array of communicative signals.

(8/10)

We believe these research axes will contribute to creating agents that not only do have human-like OOD skills, but are also able to exhibit such skills in extremely complex and realistic environments.

(9/10)

Many thanks to my brilliant coauthor and supervisor @introspection, as well as all the people who provided us with useful feedback!

(10/10)