Kazuhiro Takemoto

Hello, I'm Kazuhiro Takemoto (竹本和広), a professor at Kyushu Institute of Technology in Japan. My research combines network science and AI to understand and steer complex systems, from the dynamics of networks to the ethics and safety of large language models and the robustness of medical AI. I also apply these approaches to problems in biological, environmental, and social sciences.

Current Research

• Whispered algorithms: Steering network dynamics Adversarial perturbations as a tool for steering opinion dynamics, cooperation, and synchronization (with an analytical foundation) on networks.
• Moral code of AI: Ethics and safety of large language models How large language models make moral decisions, tested in a large-scale experiment, how such judgments scale with model size, and how safe they are when controlling robotic health attendants.
• AI's Achilles heel: Probing and protecting healthcare models The vulnerability of medical AI to universal adversarial attacks, including COVID-19 detection models, backdoor attacks, and physical camera-based attacks on skin cancer apps.