“In the historical debate within AI, 'weak methods' (general-purpose search and learning) have decisively outperformed 'strong methods' (systems imbued with specific human knowledge).”