In this episode of HD/Cast, we delve into the surprising challenges that advanced AI models face with basic grade-school math. Despite their prowess in complex tasks like writing sonnets and real-time translations, these models often stumble over simple word problems. A recent study from Apple reveals why: AI models excel at pattern matching but struggle with true logical reasoning. We explore the GSM-Symbolic benchmark and the profound limitations it exposes, discussing why AI falls short when faced with varied and unexpected math challenges. Join us as we uncover what this means for the future of AI reasoning.