AI Excels in Math but Struggles with Humanities

Astonishingly, artificial intelligence has made significant strides in Japan’s most challenging academic environments. The latest iteration of OpenAI’s model, ChatGPT 5.2 Thinking, has not only passed entrance exams for Tokyo University and Kyoto University but has outperformed human top scorers. In a closed-book, online examination conducted by LifePrompt and a renowned Japanese tutoring institution, the AI demonstrated superior performance compared to elite human candidates. However, analysis of the results revealed a surprising weakness in an area where AI scored only 25% — the essay portion of the world history section. This could indicate a key area where human students hold a competitive edge. Despite achieving a perfect score in mathematics and a 90% mark in English in tests known for their extreme difficulty, ChatGPT 5.2 Thinking faltered on a task that required structured narrative, which is currently a limitation for AI. Experts suggest that AI lacks the ability to organize complex narratives, which is a critical skill in humanities. This deficiency in long-form, coherent text highlights a current limitation of AI’s global structuring capabilities. For students preparing for university, this insight suggests a strategic direction: moving away from rote memorization and focusing on developing macro-structuring abilities, which are beyond AI’s current capabilities. The AI’s inability to manage complex, unstructured data in real-world scenarios points to areas where human skills are still invaluable. Thus, students are advised to cultivate capabilities that complement AI, such as formulating incisive questions, breaking down large tasks into AI-understandable instructions, and assembling AI-generated fragments into usable systems. In the rapidly evolving field of AI, understanding its limitations is crucial for personal development planning and career decisions.