日時：2025年6月20日(金) 13:30 ~　90分

場所：webのみ（※ 対象は特に限定しない）

講演者：Matteo Ceriscioli 先生（Oregon State University）　　　　

題目： Robust AI doesn’t just react to change, it learns the rules behind it.

概要：
What kind of internal representations enable an AI agent to remain effective when the world around it changes? It turns out that agents capable of maintaining optimal behavior across a range of shifted environments must, implicitly, learn a causal model of the world they operate in, even when their actions influence the environment. This new connection between robustness to distribution shifts and causality allows us to treat agents not only as learners and planners, but also as tools for causal discovery. Additionally, we will examine the implications of these results for transfer learning in multi-agent systems and in sequential decision tasks.

(大阪大学産業科学研究所知能推論研究分野主催のセミナーと兼ねます)

お問い合わせは以下のアドレスにメールでご連絡ください。
dser-center@biwako.shiga-u.ac.jp