AI Blunders and Human-in-the-Loop | Tomoyuki Furutani, Associate Dean of the Faculty of Policy Management
When you have been a university faculty member for a long time, you often "mess up" at work.
I messed up just the other day. During an important online meeting, a minutes-generating AI started up unintentionally. Not only did it record the entire meeting, but it also kindly summarized it and automatically sent it to all participants. Of course, the administrative staff gave me a severe scolding. "We have nothing but distrust for you, Professor." I was humiliated. I offered a profuse apology. After all, apologizing is free.
Students' failures are nothing compared to the failures of faculty members.
However, we live in an age where how you use generative AI can ruin your life.
Please be more careful than ever before.
We are in an era where it is natural to ask generative AI for everything—parenting advice, how to study, or application documents for entrance and employment exams. AI grades documents and papers generated by AI, and AI exposes vulnerabilities in human-built systems before humans even notice them. The day when children raised by AI and robots enter university is probably not that far off.
Precisely because we are an AI-premised society, the campus should provide abundant computing resources, but sadly, SFC is still lacking in terms of computing resources. I hope that someday an "Okashira" (Dean) with an understanding of large-scale funding will appear, and we can provide computing resources far superior to those of the companies where students intern. Of course, SFC faculty members have diverse views on the introduction of AI; some actively utilize AI in education and research, while others are adamantly opposed. Generative AI can even quietly teach you how to deal with the opposition.
When I entered SFC (as a 3rd-batch student) around 1992, it was an era when every student purchased (or was made to purchase) a laptop at the co-op. Remembering how that experience felt like the near future, I suggested at a meeting, "Why don't we introduce a humanoid robot or an autonomous vehicle for every student?" only to be met with fierce opposition: "Stop it, SFC will become a dystopia." A campus overflowing with robots is quite surreal to imagine. Will we eventually reach an era where robots representing students attend classes and robots representing professors teach? Perhaps we should start by letting generative AI handle the initial response at the office counter, placing dozens of robots in special classrooms, or launching our own satellite.
Intentionally incorporating human decision-making into an autonomous judgment process by AI is called "Human-in-the-Loop." Through my work, I have opportunities to be involved in exercises for such decision-making support. And I have often seen situations where the AI makes better judgments than the "Okashira." In an exercise, even if a nation or organization suffers enormous damage or a lawsuit is filed as a result of a human decision-making failure, you can just restart the exercise. In reality, however, that is not the case. One thing humans can do that AI cannot is take responsibility for decision-making failures. You can just ask generative AI how to take responsibility, too. Perhaps AI will even propose reform methods for SFC's future vision and management.
...While I was writing such nonsense, the night has fully deepened.
If I had known it would take this long, I should have had generative AI write the "Okashira Diary" for me.
On that note, an event called "SFC Tech 2026," gathering SFC's tech-oriented faculty and laboratories, will be held on the afternoon of Tuesday, November 24, 2026—the day after ORF ends at SFC—at the North Annex of the Mita Campus. My lab is currently debating whether to showcase model trains or drones. To all tech companies and organizations considering joint research with SFC or Keio University laboratories, please do come. Details will be provided separately.