Chips & Hardware
1X unveils world model to help Neo robots learn new tasks
1X has launched a world model AI designed to help its Neo humanoid robots learn new tasks from video and prompts, though its current capabilities remain limited.
Robotics company 1X has unveiled a new AI model called the 1X World Model, which is designed to help its humanoid robots learn. A world model is an artificial intelligence model that understands the dynamics of the physical world. This physics-based model uses a combination of video and prompts to allow 1X’s Neo humanoid robots to learn new tasks they were not previously trained on, according to the company.
Bernt Børnich, the founder and CEO of 1X, framed the technology as a way for robots to learn from broad datasets. “After years of developing our world model and making Neo’s design as close to human as possible, Neo can now learn from internet-scale video and apply that knowledge directly to the physical world,” Børnich said in a statement. He also claimed the system has the ability to transform any prompt into new actions, even without prior examples. However, the claim that the bot can transform any prompt into a new action is a lofty one and not entirely accurate; for instance, a user cannot tell a Neo robot to drive a car and expect it to suddenly know how to parallel park.
The release of the AI model comes as 1X is gearing up to release its Neo humanoids into the home. The company opened preorders for the humanoids in October, and a spokesperson reported that preorders exceeded expectations, though the company declined to share a specific shipping timeline or order numbers. While the company’s long-term vision is ambitious, the actual tasks the robots can currently learn from this model remain limited. A spokesperson clarified that while the world model allows the bots to attempt anything, their demonstrated capabilities are restricted to basic tasks, such as removing an air fryer basket, putting toast in a toaster, and giving a high five. 1X views the ability to learn these small tasks as a necessary first step for the robots to eventually master more complicated actions. Additionally, the world model gives the company insight into how Neo is thinking of behaving or reacting to a certain prompt, providing useful data for future training.
Why it matters
1X is preparing to bring its Neo humanoids into the home, and this world model is the foundational step toward enabling robots to learn complex, real-world tasks autonomously.