Seminar Artificial Inelligence - Dr. G.H. Lim

Title: Cognitive Robotics: From the perspective of Representation and Learning.

Abstract:

Cognitive robotics is an interdisciplinary research field with artificial intelligence (AI), machine learning, robotics, cognitive science and computational neuroscience. The ultimate goal of cognitive robotics might be to realize human-level AI for autonomous agents or robots. One of ways to achieve the goal would be to mimic human cognition which is involved in mental processes in the brain. In this talk, I’d like to share my experience of cognitive robotics from the perspective of multiscale saptio-temporal representation and learning. While most researchers in robotics focus on individual cognitive functions such as perception, manipulation, navigation, action planning, cognitive robotics is concerned with integrating those functions into a unified framework. Computation units take data in one form and change it into another form by a sequence of processes. While most processing units can perform a fixed number of operations, information which is processed by a processing unit varies and keeps changing. Especially the process of neural processing units to run deep neural networks is to propagate activation signals between connected nodes. The matter to realize human-level AI might be dependent on how to represent data or information and how to learn representations. There might be unified cognitive representations which is able to integrate multiscale cognitive functions on massive parallel computational processing units.