Sakana AI Names Schmidhuber Chief Scientific Advisor for RSI Lab
Tokyo-based Sakana AI announced on September 24, 2026 that Jürgen Schmidhuber, widely credited as a pioneer of modern deep learning and world models, will join as chief scientific advisor while keeping his other posts, Nikkei Asia confirmed the following day. Schmidhuber will help steer Sakana’s newly formed Recursive Self-Improvement Lab and its push into agent-native world models that simulate physical consequences before robots or factory systems act. The hire is cast as both a research signal and a talent magnet: Sakana argues Japan can reverse AI brain drain by pairing foundational science with its manufacturing and robotics base rather than chasing consumer chatbots alone.
Filed under Industry and dated September 26, 2026, this AI4Japan briefing treats the appointment as Japanese physical-AI research news distinct from yesterday’s workplace-robot rollouts. Schmidhuber’s public comments emphasised bridging neural architectures with robotics, citing Japanese lineage from Fukushima’s Neocognitron to Amari’s theoretical work, and framed world models as cognitive engines for monozukuri supply chains that cannot afford endless real-world trial and error.
Why it matters: Japanese factories need safer simulation before scaling autonomy. World-model research can shrink costly experiments—but only if safety cases, dataset provenance and human stop authority travel with each agent.
What it means in practice
Japanese R&D and plant leads should inventory which lines might later use simulated rollouts; demand named evaluation harnesses for RSI experiments; assign an owner for physical-AI risk reviews; run time-boxed comparisons against incumbent digital twins; and prefer designs that keep humans on irreversible actions. Anchor the hire to Sakana’s Fugu orchestration models and Japan’s AI-robot labour push.
Caveats come first. An advisory title is not a product ship date; recursive self-improvement slogans can outrun auditability; and physical AI still collides with labour and liability law. AI4Japan therefore presents the Schmidhuber appointment as directional industry context until published RSI benchmarks and factory pilots appear.
What to watch next: first public RSI Lab papers; how often Schmidhuber is in Tokyo with partners; and whether manufacturers co-fund world-model trials. Readers can continue on the AI4Japan homepage, or browse the Newsroom for additional briefings.
Bottom line: treat this update as orientation, not instruction. Japanese AI research is courting foundational talent for physical systems and remains early. Organizations that benefit most will demand simulation evidence, keep humans on shop-floor authority, and refuse to confuse an advisor announcement with finished autonomy.