Jürgen Schmidhuber Joins Sakana AI as Chief Scientific Advisor
Sakana AI announced on September 24, 2026 that Jürgen Schmidhuber, widely regarded as one of the founding figures of modern deep learning, has joined the company as Chief Scientific Advisor. He will also contribute to Sakana’s newly formed Recursive Self-Improvement (RSI) Lab while keeping his existing academic and research positions.
Details
- New role: Chief Scientific Advisor at Sakana AI, with a parallel contributing role in the company’s Recursive Self-Improvement (RSI) Lab
- Background: Schmidhuber’s foundational 1991 deep learning work underpins much of today’s AI, and his research spans World Models, Meta-Learning, and the Gödel Machine
- Stated vision: “Japan is a birthplace of foundational neural network architectures and advanced robotics… The future of intelligence is not just language; it is physical AI powered by World Models.”
- Focus areas at Sakana: Physical AI and autonomous systems that act in the real world, Agent-Native World Models for safely simulating actions before real-world deployment, and recursive self-improvement cycles for advancing machine intelligence
- Historical framing: he credited Japanese pioneers Kunihiko Fukushima (the 1979 Neocognitron) and Shun-ichi Amari, arguing Japan helped initiate the AI revolution rather than merely joining it later
- Commitment: Schmidhuber will travel to Tokyo regularly to work with Sakana’s team and support Japan’s broader AI ecosystem, alongside his current roles
What happened next
The hire follows directly from Sakana’s own research direction: the company has said Schmidhuber’s earlier work on meta-learning and recursive self-improvement already shaped two of its projects, the Darwin Gödel Machine and The AI Scientist. It also lands less than a week after Sakana publicly unveiled its Frontier Intelligence Group on September 18, a research unit dedicated to exploring biologically inspired alternatives to transformer models — giving Schmidhuber’s advisory role a clear internal home and reinforcing Sakana’s bet on long-horizon, non-mainstream AI research over near-term benchmark chasing.