Welcome!

My name is Loizos Michael, and I come from the sunny island of Cyprus, in eastern Mediterranean, Europe. I completed my undergraduate studies at University of Cyprus, with a major degree in Computer Science and a minor degree in Mathematics. I am currently pursuing a Ph.D. degree in Computer Science at Harvard University, under the supervision of Prof. Leslie Valiant (expected graduation date: May 2008). 

My main interests lie in the intersection of Artificial Intelligence and Theoretical Computer Science. I am primarily interested in the development and study of quantitative models of how humans, and biological organisms in nature more generally, solve, either in isolation or distributively, intractable every-day problems with such an apparent ease, in as-best-as-possible manner. What makes us humans the powerful computational machines we are, is neither our speed of computation, nor our infallibility, but rather our ability to reason in a way that is usually correct, and nearly optimal. Formulating theories of how humans model their world, how they acquire unaxiomatized or commonsense knowledge, what the role of learning and evolution is in achieving this task, and how this knowledge is reasoned with and manipulated, is an essential step in deriving scientifically-sound models of human intelligence. Only the development of such formal models, devoid of philosophical considerations as they are, will enable us to supply machines with human-level AI.

Perhaps two of the most fundamental aspects of human intelligence are those of learning and reasoning. It is through learning that the bulk of human knowledge is acquired, beliefs are formed, and theories of our world are developed, and it is through reasoning that such acquired knowledge, beliefs, and theories are employed to guide our decision making mechanisms. An understanding of the principles underlying the learning and reasoning phenomena is of primary importance if we are to replicate the intricacies of human intelligence. My current research focuses on the formalization of certain aspects of autodidactic learning and reasoning, the processes by which humans learn and employ knowledge in the absence of teacher-prepared material. Within this context I am investigating the computational feasibility of learning, and the role of reasoning in drawing conclusions through learned, or otherwise acquired rules, as well as the interaction of the learning and reasoning processes. The theoretical insights gained from this study have been used in the development of a system for the automated and massive-scale extraction of commonsense knowledge encoded in text.

Specific Areas of Interest

Reasoning about Actions and Change, Non-Monotonic and Default Reasoning, Computational Learning Theory, Computational Evolution Theory, Artificial Life, Distributed Computation, Game Theory.