³Actually, Deep Blue mentioned in Chapter 3 did already use some learning as well: It analysed data from several chess databases, including 700,000 historical games played by human grandmasters, to compute another heuristic.