By understanding more about how the brain processes information, techniques from neuroscience, combined with physics of information, can help us translate human thinking from its biology into fundamental mathematical functions of intelligence.
Moving on from deep learning, designing new AI paradigms will require better mathematics for learning in itself. We start from understanding the emergence of goals in agents, eventually aiming at building a global framework for the problem of credit assignment.
We work towards the design of societies of AI agents that can truly parallelize their computation, learn collectively from each other, and invent new mechanisms that work on a collective scale, in order to work on a problem together.