Artificial intelligence (AI) is gradually changing the landscape of cancer pathology diagnosis. With recent progress in weakly-supervised machine learning algorithms, explainable AI, and high-resolution whole-slide digitization, AI methods have the potential for augmenting cancer subtyping and personalizing treatments. In this talk, I will describe interpretable and generalizable methods for extracting biological signals from gigapixel whole-slide pathology images, present case studies on integrative pathology-multiomics analyses, and identify the challenges for further progress in automated pathology evaluation.