Yuta Sugafuji: Auxiliary Lines Connecting Education and AI

While working as a CNS consultant at the Media Center, gazing at Gulliver Pond (Kamoike) at SFC through the glass, I frequently witnessed moments where the same information could change a person's understanding and their next step depending on how it was conveyed. Listening to their troubles, rephrasing, organizing the sequence, and adding context and options—that alone was enough to get someone who had stalled moving again. While being exposed to the philosophy of the Internet at the Jun Murai Laboratory and being involved in natural language processing research, I realized that what moves society is not technology itself, but the "words" used to deliver technology into people's hands—meaning-making, translation, and the design of learning.

Learning is the act of turning "not understanding" into "understanding." It involves setting definitions, posing questions, supporting with analogies, and nipping misunderstandings in the bud. Generative AI also draws out knowledge, assists reasoning, and creates expression through the medium of language. That is why I have continued a process of trial and error to find the "shortest distance to understanding" by adjusting how premises are set and the granularity of questions. When words change, understanding changes, and the number of futures one can choose increases.

Starting from this perspective, I co-founded a company with a friend. We built a system where online tutors and students meet. The goal was not just to gather good teachers, but to provide an entrance where people who stumble before they even start learning can take their first step.

On our company's YouTube channel, "Manabi Square," we start by explaining "why we think this way" regarding complex geometric figures. The number of subscribers is approximately 84,000, and we have covered nearly 1,000 problems. I even received a postcard from a viewer saying, "I watched because it was your explanation," which taught me that people seek a companion in thinking rather than just the correct answer. At the end of last year, I published and reprinted a book titled "The Book That Lets Anyone Solve Kaisei and Nada Junior High School Math Entrance Exam Problems in Just One Day," delivering the sensation that a single auxiliary line can change the world, even through paper media.

Currently, I am scaling that companionship through technology. I have been involved in machine learning for about 10 years since my student days, working on generative AI product development and promotion at DeNA, and handling implementation support and workshops at my own company. I have picked up voices from the field and accompanied them until the technology took root through the twin wheels of product and operation. Generative AI only demonstrates its power when one verbalizes and poses appropriate questions. I believe the core is not "answering on behalf of humans," but "refining questions into a form that allows humans to keep thinking." I will continue to move back and forth between education and generative AI, drawing "auxiliary lines" that widen the entrance to learning so that everyone can proceed without hesitation.