3.0

credits

Average Course Rating

This class introduces a model of natural intelligence using the formalism developed in the field of artificial intelligence. We start with the simplest possible model, that an intelligence desires to decide among several options and select the 'best' one, and write it down formally as an optimization problem. We then add increasing degrees of difficulty and uncertainty, starting with epistemic uncertainty, where we introduce probability theory. And then adding aleatoric uncertainty, where we introduce statistical learning theory. As the intelligence faces increasingly complicated scenarios, we introduce increasingly complicated theoretical and pragmatic solutions, covering the bias/variance trade-off, model selection, robustness to adversarial events, control theory, and finally dynamics and dependence. Each week will have a reading assignment of a technical paper as well as a blog post, and potentially some math/programming assignments. In our once a week mandatory class meetings, we will discuss any sources of confusion, be they technical or philosophical.