I'd strongly recommend these, if you are interested in these kinds of things. I look forward eagerly to watching the annual Christmas Tree lecture every year (this year on on Bayesian trees and BDDs, to be held December 8, probably posted online a few days later).

One year Knuth offhandedly mentioned a question I had posed in an email a few days earlier (regarding TAOCP), which was a big thrill, and indicates that despite the warning that email may take 6 months to reach him, it is often much quicker.