Tokyo-based artificial intelligence company Sakana AI has made a significant move in the AI landscape by appointing Jürgen Schmidhuber, widely recognized as the "father of modern AI," as its Chief Scientific Advisor. Schmidhuber, who pioneered deep learning and world models, is set to lead Sakana's new Recursive Self-Improvement (RSI) Lab, a research initiative focused on creating AI systems capable of continuously enhancing their own capabilities.
Building the Future of AI Self-Improvement
Sakana AI's decision to bring Schmidhuber on board underscores its ambition to push the boundaries of AI autonomy. The company has already begun implementing Schmidhuber's theoretical concepts, such as the Darwin Gödel Machine, which he first proposed in the 1990s. This machine is designed to improve its own code and logic recursively, a concept that could be pivotal in the development of next-generation AI systems.
The RSI Lab’s work is particularly relevant as the AI industry grapples with questions of scalability, adaptability, and long-term intelligence. Schmidhuber's involvement signals that Sakana AI is not merely chasing current trends but is investing in foundational research that could shape the next wave of AI advancements—potentially even influencing the evolution of systems like ChatGPT.
Legacy and Vision for the Future
With a career spanning decades and contributions that include the invention of LSTM (Long Short-Term Memory) networks and the development of world models, Schmidhuber has long been a visionary in the field. His move to Sakana AI is a strong indicator of the company's commitment to innovation and scientific rigor. As AI systems become more complex, the need for self-improving architectures becomes paramount, and Schmidhuber's expertise could be instrumental in realizing that vision.
With this strategic hire, Sakana AI positions itself at the forefront of AI research, aiming to develop systems that don’t just learn from data but also evolve and improve on their own. It’s a bold step toward a future where AI systems may be capable of advancing beyond their original programming.