ChatGPT Tops University of Tokyo Entrance Exam After Failing Two Years Ago [Current Affairs Show]

Scored 50 Points Higher Than the Top Human Applicant
Shockwaves in Japan’s Education Sector... Job Threats for University Graduates

■ Broadcast: The Asia Business Daily's "So Jongseop's Current Affairs Show"

■ Host: So Jongseop, Political Specialist

■ Director: Lee Miri, Producer

■ Guest: Reporter Lee Hyunwoo

ChatGPT took the entrance exam for the University of Tokyo, considered the most competitive university entrance exam in Japan, and not only passed, but scored 50 points higher than the top human test-taker. This is a dramatic turnaround compared to just two years ago, when it failed the same exam. While some are praising this as evidence of remarkable advances in AI technology within a short period, there is also growing concern among students preparing for entrance exams and recent graduates that AI is becoming a realistic threat to jobs.

ChatGPT Tops University of Tokyo Medical School Entrance Exam... Dramatic Improvement in Two Years

Reuters Yonhap News

This time, ChatGPT scored 503.59 out of a perfect 550 points on the University of Tokyo medical school exam. This is about 50 points higher than the top score among this year's admitted students, which was 453.60 according to the university's announcement. It is reported that this score would also make ChatGPT the top scorer in the law school entrance exam. This exam was conducted by Japanese AI venture company LifePrompt, which had three major generative AIs—OpenAI's ChatGPT-5.2 Thinking, Google's Gemini 3 Pro Preview, and Anthropic's Claude 4.5 Opus—solve the actual 2026 University of Tokyo entrance exam questions. While Gemini and Claude also passed, ChatGPT received the highest score and drew the most attention.

To ensure fairness, ChatGPT was blocked from internet searches and relied solely on its internal data for the test. By subject, it achieved the highest scores in mathematics, earning perfect marks in all math-related questions, regardless of whether they were in liberal arts or sciences. However, it still showed limitations in essay-type questions for history subjects such as world history and Japanese history. Weaknesses were observed in selecting and organizing information, constructing sentences, and understanding the intent of the questions. The nature of the test, which included many tricky questions, made it difficult for ChatGPT to fully grasp the intent of some prompts.

The main reasons ChatGPT was able to go from failing to becoming the top scorer in just two years were a significant improvement in computational processing power and a dramatic increase in input data volume. Some experts predict that, at this rate, it will be able to achieve perfect scores in almost all subjects within the next two years. There are also expectations that its performance in history essay questions, which was identified as a current limitation, will be significantly enhanced.

Japan's Education Sector in Shock... Job Threats for New University Graduates

Reuters Yonhap News

This result has sent shockwaves through Japanese society. The University of Tokyo's medical and law schools have traditionally been considered the domain of only the very top students, but with AI easily achieving the highest scores, both students and teachers are experiencing a sense of upheaval. The core issue is how to educate children going forward.

Until now, most countries have operated education systems primarily focused on rote memorization, with exams designed to test memorization ability. However, since modern AI is capable of collecting and processing all text-based information, it is now virtually impossible for humans to memorize more than AI, no matter how hard they try. As a result, there is a growing movement calling for a complete overhaul of education systems that rely on rote learning.

This issue is directly tied to employment prospects. In many countries, including the United States, new hiring has already been drastically reduced, and simple work that was traditionally assigned to new employees is increasingly being handled by AI, making human workers unnecessary. Graduates who have completed rote-based education cannot outperform AI in intelligence, so from a corporate perspective, it is much more advantageous to use AI. As a result, it is becoming increasingly difficult for university graduates and new employees without experience to secure jobs.

This problem is expected to be even more pronounced in white-collar occupations. While current generative AI mainly serves as a work assistance tool, the next generation of agentic AI refers to AI capable of independently carrying out all types of office work. Until now, employees would ask ChatGPT questions, request code, or perform searches, but in the era of agentic AI, the AI itself will generate the questions, hold meetings with other AIs, share opinions, and complete entire projects on its own. It is predicted that within five years, a structure in which a single manager collaborates with agentic AIs to handle nearly all tasks will become standard in most companies.

Even blue-collar jobs are not entirely safe. Robots such as Hyundai Motor's Atlas are already being deployed in factories, and although it will take more time to develop fully autonomous, self-thinking AI optimized for robots, blue-collar fields are considered to have more leeway than white-collar ones. However, as the supply chain for android-type robot manufacturing expands and more power plants are built to support the enormous energy demands of such operations, this sector is also expected to face significant threats.

Exponential Acceleration of AI Development... Calls for Enhanced Entrepreneurship and Investment Education

Reuters Yonhap News

The pace of change is also much faster than initially anticipated. Until recently, it was believed that it would take 10 to 15 years for agentic AI and advanced robotics to become widespread, but now that AI can collect its own data and set and solve its own tasks, the acceleration is evident. Many technology development companies have already built dedicated messaging platforms exclusively for AI-to-AI communication. As AIs handle work through direct communication without human intervention, the pace of progress has accelerated dramatically.

Against this backdrop, Wall Street is seeing a trend where a single CEO can create a company worth hundreds of billions of won—so-called "one-person unicorns." Tasks that once required hundreds of people can now be accomplished by just one person. Experts predict that unless students preparing for employment develop their own entrepreneurial ideas and pursue independence, it will become nearly impossible to join any organization at all.

Accordingly, there are growing calls for a fundamental shift in education. The traditional approach of preparing for stable office jobs through rote learning is no longer valid in an era when the value of labor is rapidly diminishing. Instead, there is a rising consensus that entrepreneurship and financial education—which were previously marginalized—should be significantly strengthened, focusing on how to launch a business or generate income through investment.

This content was produced with the assistance of AI translation services.