Neo humanoid maker 1X releases world model to help bots learn what they see

1X, the innovative robotics firm behind the Neo humanoid robot, has introduced a groundbreaking AI model designed to help robots comprehend and navigate the complexities of the real world. This new physics-based model, termed the 1X World Model, harnesses the power of video and prompts to equip Neo robots with advanced learning capabilities. According to 1X, this model allows Neo robots to acquire new skills beyond their initial training. By utilizing video data as a learning tool, these humanoids can be programmed to tackle tasks they have never encountered before. As 1X prepares to launch its Neo humanoids for home use, pre-orders began in October, with the company expressing optimism about shipping the robots within the year. However, details regarding the shipment timeline and the number of pre-orders remain undisclosed, aside from indications that interest has surpassed expectations. "After years of refining our world model and striving to make Neo's design as human-like as possible, we are thrilled to enable Neo to learn from a vast array of online video content and apply that knowledge to the real world," stated Bernt Børnich, founder and CEO of 1X. He emphasized that this capability allows Neo to convert any prompt into actionable responses, potentially paving the way for self-teaching abilities in robots. While the claim that Neo can transform any prompt into a new action is ambitious, it’s important to clarify that the robots cannot instantly master complex tasks, such as driving a car, merely from video prompts. A spokesperson from 1X clarified that the current model does not allow Neo to learn a new task immediately; rather, it processes video data associated with specific prompts to enhance the overall understanding of the physical environment among all Neo robots. This advanced learning mechanism also provides users with insights into Neo's decision-making processes, offering a glimpse into how the robot might respond to various prompts. Such behavioral insights could be pivotal as 1X continues to refine its models, ultimately aiming to equip robots with the ability to react to unfamiliar prompts effectively.