World models are about to reach robots and cars
Wired says AI that learns from visual simulations of the real world could shape home robots and driverless cars as soon as 2027.
The facts
- According to Wired, world models take in visual inputs and generate 3D environments that simulate physics, unlike language models, which are text-based.
- Niantic Spatial's CTO Brian McClendon says photos of a robot's intended workspace could be turned into a 3D model and used to run "millions of simulations".
- Wired reports that Wayve builds a world model partly from dashcam footage to recreate hazardous roads that cannot safely be tested in real life.
- Wired says Yann LeCun's startup raised over a billion dollars, and AMD bought Fei-Fei Li's World Labs this year for over $8 billion.
Case notes · Danny's take
Why it matters
Language models learnt from text. World models learn from footage of the physical world, and someone has to record that footage. Wired's writer sold videos of himself doing chores for what he calls a measly payout. That's the part most people miss: the training data is people's daily lives, and the pay may not match its value. Whoever owns the footage owns the advantage.
For traders
Watch where the money actually lands: funding rounds, acquisitions and compute spending tied to world models. Big deals like the one reported for World Labs show where the narrative is, not whether the products work. Ask who is selling to you.
For builders
Real-world video and 3D scans are becoming valuable training data. If you build in robotics, mapping or gaming, think about how you collect that data and how you pay the people who supply it. Simulation-based testing is also a cheaper way to try risky scenarios.
Not financial advice. Do your own investigation.
This story was summarised with AI from Wired's reporting and checked against our editorial policy. Danny's take reflects his views. Disclaimer.
Danny Walter
Founder of Moon Republic, Volt Terminal and Moon Republic Academy. Author of The Market Is a Crime Scene and After the Applause.