Startup bets video game data can train physical AI foundation models

It is a startup focused on developing foundation models for physical AI using video game data to train robots.

Instead of collecting massive amounts of real-world data (costly and slow), the company uses data from video games that simulate complex physical interactions.

Yes, it could benefit sectors like manufacturing, logistics, and healthcare in the region by enabling faster and cheaper robot training.

The main challenge is ensuring that video game data accurately reflects real-world physics and interactions to avoid errors when deploying models on actual robots.