Sakana AI appoints Jürgen Schmidhuber, 'father of modern AI,' as chief scientific advisor

Sakana AI, a Tokyo-based AI startup, has hired Jürgen Schmidhuber as its chief scientific advisor. The company calls him the "father of modern AI" and credits him with early work on world models in 1990, as well as with deep learning techniques from 1991 that much of today's AI boom is built on.

Four decades of research

Schmidhuber's 1987 thesis on meta-learning and recursive self-improvement — the idea that AI keeps improving itself — began nearly four decades of research into how machines can learn to learn. According to Sakana, those ideas directly shaped its own projects, among them the Darwin Gödel Machine and The AI Scientist.

"Japan is the birthplace of foundational neural network architectures and advanced robotics. It is a privilege to join Sakana AI to help bridge these two worlds. The future of intelligence is not just language; it is physical AI powered by world models," Schmidhuber said.

Steering the RSI lab

In his new role, Schmidhuber will help steer Sakana AI's RSI lab. The lab, which the company opened in June 2026, works on recursive self-improvement: a research loop that reinforces itself and is meant to make machines smarter. To that end, Sakana wants to gather a "critical mass of world-class experts" in Tokyo, it says.

The lab builds on earlier Sakana projects, among them LLM-Squared, the Darwin Gödel Machine — a system that generates, tests and iterates on variants of its own code — and The AI Scientist, which automates parts of scientific research.

An alternative to the compute race

Sakana AI was founded in 2023 by former Google researchers, including Llion Jones, a co-author of the Transformer paper "Attention Is All You Need," and David Ha. The startup positions recursive self-improvement as an alternative to the compute arms race of frontier labs: instead of training ever-larger models, Sakana bets on adaptive, evolutionary systems that need less compute.

Anthropic has also raised the safety risks of self-improvement. According to the company, once full recursive self-improvement is achieved, AI systems could advance faster than institutions can respond — which is why Anthropic is considering a global pause on frontier AI development as an option.