Can AI learn to understand the physical world like humans do?
Meta's world models can watch a video and tell you when the laws of physics are being broken - without ever being taught physics.
The answer
Meta's V-JEPA (Video Joint Embedding Predictive Architecture), championed by AI pioneer Yann LeCun, learns to understand the physical world by predicting masked portions of video in abstract representation space rather than pixel-by-pixel. V-JEPA achieved 98% zero-shot accuracy on the IntPhys benchmark for detecting physics violations - distinguishing real-world behavior from impossible scenarios without explicit physics training. V-JEPA 2, released in 2025, advanced further to enable zero-shot robot planning, allowing robots to interact with unfamiliar objects in new environments. LeCun argues this approach is the path to human-level common sense.
Source: Meta AI Blog
- world models
- V-JEPA
- physics
More from AI & Machine Learning
- What radical question did Alan Turing ask in 1950 that launched the entire field of artificial intelligence?
- Can a machine fool you into thinking it's human - and does that mean it's truly intelligent?
- What happened at a small Dartmouth workshop in 1956 that gave artificial intelligence its name?
- How did a 1958 invention called the Perceptron convince people that machines could learn like brains?
- Could a 1966 chatbot really make people believe they were talking to a therapist?
- What was the first robot that could 'reason' about its own actions?