Sakana AI Hires Jürgen Schmidhuber as Chief Advisor

Tokyo-based artificial intelligence startup Sakana AI has appointed Jürgen Schmidhuber as its new chief scientific advisor. In this role, the prominent computer scientist will help guide the company's newly established Recursive Self-Improvement (RSI) laboratory. The initiative aims to assemble a concentrated group of global experts in Japan to develop autonomous systems capable of conducting scientific research and improving their own capabilities over time.

Schmidhuber is widely recognized for his foundational contributions to the field of machine learning. Sakana AI noted that his 1987 thesis on meta-learning laid the groundwork for machines that learn how to learn. Furthermore, his work on world models in 1990 and deep learning architectures in 1991 continues to underpin modern generative AI systems. The startup has already drawn heavily from these concepts, citing his historical research as a direct inspiration for its own projects, including the Darwin Gödel Machine and its automated research agent, The AI Scientist.

The appointment highlights a strategic shift toward physical AI, which combines robotics with advanced neural networks. Schmidhuber expressed his intention to bridge Japan's history in robotics with modern AI, stating that the future of the field lies in "physical AI powered by world models" rather than just language processing. By establishing the RSI lab in Tokyo, Sakana AI hopes to build systems that can simulate and interact with the physical world.

For AI practitioners and developers, this high-profile hiring signals a serious industry push toward autonomous, self-correcting software. Instead of relying solely on human engineers to fine-tune models, future workflows may increasingly leverage recursive self-improvement loops. If Sakana AI successfully commercializes Schmidhuber's theories, developers could soon gain access to tools that autonomously debug, optimize, and upgrade themselves, shifting the practitioner's role from active coder to high-level system supervisor.