Odyssey-3: A World Model That Teaches Robots Physics Through Video

Odyssey has unveiled Odyssey-3, an innovative world model designed to teach robots physical laws by analyzing video data. Unlike traditional methods requiring thousands of hours of specialized training, Odyssey-3 mimics the way a child learns by 'absorbing' physical principles from observing objects. This approach significantly reduces the time needed to adapt robots to specific tasks, requiring only a few dozen hours of data to configure manipulators or mobility systems. On October 8, the company released a research preview of the model. The article on Habr provides a detailed breakdown of the Odyssey-3 architecture, its training methods, and an analysis of its record-breaking performance metrics. The author critically evaluates the technology, distinguishing between genuine scientific breakthroughs and marketing hype, while exploring the potential for this model to create more flexible and autonomous robotic systems.
This is a summary. Read the full article at the original source:
HabrRelated stories
How Google Taught a Tiny Model to Understand Text, Images, Video, and Audio: A Deep Dive into EmbeddingGemma 2
On October 6, 2026, Google DeepMind released EmbeddingGemma 2, a compact multimodal embedding model. The model maps text, code, images, video, and aud…
Satya Nadella says we should assume all AI models are ‘compromised’
Microsoft CEO Satya Nadella has issued a stark warning regarding the security of advanced artificial intelligence models. In a recent statement, Nadel…
Neural network by recipe: calculating weights instead of storing matrices
This Habr article explores alternative methods for storing neural network parameters. Instead of the traditional approach of storing massive weight ma…



