Note · 1 min read
Startup licenses video game data to train AI world models
British startup Worldmodeldata is turning video game footage and player inputs into training data for AI systems that need to act in the physical world. The company says it has licensed nearly 1 million hours of game data.
Oossa · About Oossa
Worldmodeldata, a British startup advised by AI researcher Yann LeCun, is gathering visual game footage and the button presses and controller movements that go with it. It plans to package the data for labs building “world models”—AI designed to understand how actions affect the physical world.
The company says it has licensed nearly 1 million hours of data from video game studios, but has not named them. The idea is that varied game situations could help train systems for tasks such as controlling robots or vehicles, though researchers have yet to prove that game data will transfer well to real-world work.
Why it matters
If game data can help AI learn how actions change its surroundings, it could offer researchers a much larger source of training examples than hands-on demonstrations alone.
Sources & references
| # | Source | Outlet | Date | Key takeaway |
|---|---|---|---|---|
| 1 | The Next Evolution of AI Is Learning From Your Dodgy Gaming Skills ↗ | Wired | Sep 28, 2026 | A British startup is shaping video game inputs into training data for AI models that can navigate the physical world. |
1 sources
Last updated:
Oossa · Newsletter
The week in AI, explained
Every Monday: the stories worth knowing, in plain language. Free, no spam.