RIKEN highlights brain-inspired AI in new feature on Amari’s work

This RIKEN feature is not a product launch or a funding round, but it’s a useful reminder of how deeply today’s AI stack is rooted in decades of theoretical work. Shun’ichi Amari’s contributions—independent component analysis, natural gradient methods, and the broader lens of information geometry—are part of the mathematical toolkit that makes training deep networks tractable at scale. By explicitly framing brain science and AI as “two intelligence systems” that should co‑evolve, RIKEN is arguing against the idea that AGI can be pursued purely as an engineering race divorced from neuroscience.

In practical terms, the piece signals continued institutional support in Japan for cross‑disciplinary work that links cognition, learning theory and machine learning. That matters because some of the hard open problems on the way to AGI—robust generalisation, continual learning, interpretability—may end up borrowing as much from biological systems as from scaling laws. RIKEN’s message is that investing in that theoretical bridge is a strategic asset, not an academic luxury, for a country that wants to stay relevant in the next era of AI.