I am a PhD student in the Commonsense Computing, Software Agents and Society of Mind groups at the Media Lab. My advisors are Henry Lieberman and Marvin Minsky. My ultimate goal is to build non-human computers that can learn and think about the world as resourcefully as their creators. My research interests include spatial reasoning, cognitive architectures, and commonsense {knowledge representation, knowledge acquisition, reasoning}. My investigations span the fields of artificial intelligence, computer science, cognitive science, neuroscience, linguistics and philosophy. My current research focuses on these problems:
Context-dependent category construction: Trying to understand the nature of categories in insolation of problem-solving is like trying to understand a single word without reading the rest of the sentence. A major problem in a cognitive architecture is retrieving and integrating knowledge from many sources. Knowledge comes from the environment, the agent's goals, background commonsense and procedural knowledge. In my MS thesis, I am focusing on ways to construct concept descriptions within a problem-solving context.
Overcoming brittleness in AI systems: The key to human level resourcefulness involves having many ways to solve, represent and learn through problem solving. The Emotion Machine is an architecture that addresses all of these issues, namely selecting many "ways to think", connecting analogous representations via the Panalogy representation, and keeping traces of previous decisions and using reflective processes to improve the effectiveness of credit assignment.
|
