Sakana AI has appointed Jürgen Schmidhuber as Chief Scientific Advisor. The Tokyo-based startup cites his 1991 work on deep learning and earlier research into world models as foundational to its own projects. Schmidhuber, often described as the father of modern AI, developed the concepts behind recursive self-improvement in his 1987 thesis. These ideas directly inform Sakana’s current initiatives, including the Darwin Gödel Machine and The AI Scientist. The new role will allow him to guide the company’s RSI lab, which aims to create a self-reinforcing research loop that makes machines smarter over time.
This appointment signals a strategic pivot away from pure language models toward physical AI capable of simulating and interacting with the real world. Schmidhuber argues that future intelligence relies on world models rather than just text processing. By bringing his expertise to Tokyo, Sakana intends to reunite Japan’s legacy in robotics with modern neural network architectures. The goal is to gather a critical mass of experts to build autonomous systems that understand physical environments.
* Schmidhuber will lead the new RSI lab focused on recursive self-improvement
* The team aims to combine Japanese robotics heritage with deep learning
* Projects include the Darwin Gödel Machine and The AI Scientist
