An Information-Theoretic Explanation for the Adversarial Fragility of AI Classifiers | Read Paper on Bytez