This episode explores a new idea emerging from DeNA’s AI strategy: building real-world environments—almost like entire mini cities—where AI-powered robots can continuously learn and operate.
While most AI so far has lived inside screens, physical AI requires interaction with real space. That makes the question of “where to train” increasingly important.
Instead of limited lab setups, the concept is to create large-scale, realistic environments—combining warehouses, offices, retail spaces, and more—where robots can operate, learn from failure, and improve over time.
This shift is less about a single technology and more about infrastructure. Just as the internet required data centers, physical AI may require dedicated real-world environments for learning.
Over time, these spaces could expand beyond individual buildings into larger areas, potentially reshaping how cities are designed.
It may be an early sign of a broader transition—from a world where robots exist, to one where they are part of the underlying assumptions of everyday life.