Abstract:  In this talk I will outline one of the long-term goals of evolutionary robotics that has philosophical appeal: how can the abstract aspects of cognition (such as categorization, self-awareness, or language) ultimately be grounded in the tangible dynamics of brain-body-environment interaction dynamics?

In the first third of my talk I will explain the standard methodology of evolutionary robotics, and place it in the larger context of embodied cognition research. In the middle third, I will describe a series of experiments that ground some of these cognitive building blocks in perception and action. I will extrapolate these works to argue that evolutionary robotics is a good candidate methodology for creating a smooth gradient along which we can engineer incrementally more cognitive machines. In the final third, I will propose how this approach may provide a solution to a major growing ethical concern, which is how to create superintelligent yet human-friendly machines.