Sakana AI hires Jürgen Schmidhuber as Chief Scientific Advisor

By Splitfeed EditorialLast updated Sep 26, 2026, 19:17 UTC

On September 24, 2025, Tokyo-based Sakana AI announced that Jürgen Schmidhuber would join the company as Chief Scientific Advisor. Schmidhuber is a researcher whose work in the 1990s contributed foundational ideas to deep learning and world models, and Sakana has described him as the father of modern AI. Alongside the hire, Sakana formally established an RSI Lab dedicated to recursive self-improvement research, a field centered on building AI systems capable of iteratively modifying and advancing themselves. Schmidhuber's influence on Sakana's work was already present before this appointment, with his ideas having shaped the company's Darwin Gödel Machine project.

The hire matters because it signals that Sakana is moving recursive self-improvement from a theoretical interest into a named institutional priority. Recursive self-improvement sits at the center of longstanding debates about AI development trajectories: optimists see it as a path to rapid capability gains that could accelerate beneficial applications, while critics argue it raises difficult questions about control, predictability, and the risks of systems that change themselves in ways their designers cannot fully anticipate. Attaching a figure of Schmidhuber's stature to that research program lends it credibility and visibility, and the creation of a dedicated lab suggests the work will receive sustained organizational attention rather than being incidental to other projects.

None of the three camps have published reactions to this story yet. The Pro-AI camp would typically welcome the hire as evidence that serious scientific talent is committed to pushing AI capabilities forward in a structured research environment. The Anti-AI camp would likely point to the recursive self-improvement focus itself as cause for concern, arguing that formalizing research into self-modifying systems accelerates risks that are not yet well understood. The Middle Ground camp would probably acknowledge Schmidhuber's contributions while calling for transparency about the RSI Lab's safety practices and governance frameworks before drawing broader conclusions.

The details worth watching include any publications or technical disclosures from the RSI Lab, further announcements about the lab's scope, staffing, or funding, and whether Sakana pairs this research direction with explicit safety commitments. The Darwin Gödel Machine project may also serve as an early indicator of the direction the RSI Lab intends to take.