Ming Hay Chung, mhc48@cam.ac.uk
United Kingdom/China
Engineering, Sidney Sussex College
PhD thesis: Learning to Plan in Model-based Reinforcement Learning.
Research interests:
1. Reinforcement Learning.
2. Deep learning.
3. Theoretical neuroscience.
4. AI alignment.
My PhD focuses on building artificial intelligence that can reason and plan like humans. When faced with an unfamiliar situation, we may think of several possible actions and simulate the corresponding future (e.g., what will happen if I hit the tennis from the left and right angle?), thereby allowing us to choose the action with the optimal result. However, we may only rely on our habits without overthinking when faced with a familiar situation such as driving home. This distinction demonstrates that planning should be a flexible and learnable process instead of a fixed process. My PhD thesis will investigate how to build artificial intelligence that learns this planning process by interacting with the environment and how such learnable self-interaction may possibly yield more powerful cognitive capabilities such as reasoning, dreaming and thinking.
Who or what inspired you to pursue your research interests?
I have always been fascinated by how human learns and adapts to diverse environments. It is a miracle that a single neuron, with simple learning rules, when grouped together with other similar neurons, can give rise to such sophisticated intelligence. Is it possible to build artificial intelligence that borrows ideas underlying this miracle? To answer this question, I have been searching for biologically plausible learning rules and testing them in artificial intelligence. I am also interested in the collective behaviour of neurons and how high-level cognitive abilities such as planning can arise in our brains.