Hi! I’m a PhD student at Georgia Tech trying to figure out how to make robots do all my work. My research is on Reinforcement Learning because it is the most promising framework for building general intelligence. In my view, reinforcement learning can bind unsupervised and supervised learning in a way that is useful for taking actions in a real environment.
These days, I am working on a small piece of intelligence, compositionality and curriculum. The idea is that humans never learn very hard tasks from scratch. We have years to learn simple skills and build them up on top of each other to solve increasingly complex tasks. I want machines to learn to do the same.
My advisor at Georgia Tech is Dr. Charles Isbell
Oh, you’re still here? Ok, I also strongly believe humanity has dwelled long enough on one planet in a universe of trillions and needs to move out of its parents’ basement! We have to leave if we ever want to answer the hard questions such as “what the heck is going on out there?” and “what’s up with life anyways? Is it special or pretty normal? lol”. Unfortunately, our bodies have too much water and too little carbon fibre. We need something that is more suitable to harsh conditions in space but just as smart as us (or even more!). Robots are ideal for exploration and setting up a base before humans arrive. But they need to be more autonomous than they are because of the long distances and time delays involved. In order to build complicated structures efficiently, robots need to be able to work with minimal human supervision, perhaps an instruction, maybe in the form of a reward or demonstration, every radio relay (15-20 mins for Mars).