Sakana AI, the Japanese artificial intelligence startup known for its work on nature-inspired algorithms, has hired Jürgen Schmidhuber, a prominent researcher widely credited with foundational contributions to deep learning and world models.
What Happened
The move marks a significant acquisition of talent for Sakana AI, bringing in one of the most cited scientists in the field of artificial intelligence. Schmidhuber is historically recognized for his pioneering work in recurrent neural networks, specifically the Long Short-Term Memory (LSTM) architecture, which became a standard for sequence modeling. He is also associated with the concept of "world models," which involve agents learning to predict their environment to plan actions.
Why It Matters
For Sakana AI, the addition of Schmidhuber signals a strategic effort to blend its existing focus on evolutionary and nature-inspired approaches with the theoretical rigor of deep learning history. As the industry moves toward more autonomous agents and complex reasoning systems, Schmidhuber’s expertise in predictive models and curiosity-driven learning offers a distinct theoretical framework that differs from the scaling laws currently dominating the field. This hire may influence how Japanese AI labs position themselves in the global race for AGI, emphasizing architectural innovation over brute-force compute.
The Bottom Line
Sakana AI has secured a major theoretical asset in Jürgen Schmidhuber, potentially reshaping its research direction toward advanced world models and autonomous agent capabilities.