Artificial Intelligence is the study of problems so ill-defined that the algorithmics people put us in our proper place when we asked them.
Machine Learning is when you assume there is certain structure in the data and then you discover it.
PCA is a method of constructing features from the noisiest input dimensions by means of a linear projection.
Genetic Programming works well when your evaluation function knows exactly what the solution is.
Central Limit Theorem tells us that with increasing number of attempts we achieve mediocrity almost surely and this is normal.
Hypothesis testing theory is a means of figuring out how many paired trials we need to reject the null hypothesis.
Powerful novel technique : an algorithm for which we could find a dataset on which it beats the obvious alternative.