Nobody knows the trajectory of AI, it's a genie unleashed. Which enterprise; small or big, academic, research or commercial will bring what successful AI innovation to the table also remains unknown. Norwegian startup 1X has unveiled its "1X World Model," a physics-based AI system that trains Neo humanoid robots by watching videos paired with simple prompts.
It captures the data, and shares learned behaviors across its network of bots. This mimics human learning from demonstration, marking a shift from scripted instructions to adaptive observation. Current demos show Neo mastering basics such as removing air fryer trays and making toast, but 1X claims it can "transform any prompt into new actions." The model processes real-world physics from video inputs, enabling robots to generalize skills without exhaustive human intervention.
As 1X prepares to ship Neo to preorder customers, this positions the company alongside rivals like Skild AI and Field AI in the race for observation-based robotics. Achieving this involved building a distributed learning system where video data from one Neo feeds the entire fleet, accelerating skill acquisition through collective experience. While tasks remain simple, the innovation lies in its scalability.
This marks the beginning of video-driven world modeling — leveraging advances in AI vision and simulation to bridge digital training with physical execution. 1X's Norwegian roots and focus on home robots highlight how nimble startups are pioneering this edge. This opens floodgates to a robotic revolution, supercharging the AI era by flooding physical spaces with self-improving machines that learn on the fly.
WATCH ROBOTS RISE: VIDEO LEARNING USHERS IN THE ERA OF AUTONOMOUS PHYSICAL AI!