12.11.2015: Daan Wierstra

How to Build an Artificial General Intelligence

Daan Wierstra, Google DeepMind

 

Abstract:

At Google DeepMind, we focus on developing general learning algorithms to push the frontier of artificial general intelligence (AGI) research. Endeavouring to eventually construct a full AGI, agent architecture design and the non-trivial matter of `how to put it all together' figures prominently in our mission. As such, DeepMind operates at the crossroads between science and engineering. The first part of this talk will provide an overview of some of our recent research, ranging from advances in (deep) reinforcement learning and (deep) variational inference methods to (deep) memory techniques. In the second part of this talk, I will elaborate on some of our general design philosophy, confront some of the profound dilemmas inherited from disparate fields such as machine learning and computational neuroscience, and explain why a great deal of work might still be ahead of us.