An OpenAI-backed humanoid robot startup says it's moving away from using humans to train its Optimus rival

In a groundbreaking shift within the robotics industry, 1X, a startup backed by OpenAI, has unveiled a new AI model that promises to reduce reliance on human operators for training its humanoid robot, Neo. This development, announced by CEO Bernt Børnich, is set to transform how humanoid robots learn to navigate their environments by allowing them to learn directly from their own experiences captured on video. Traditionally, humanoid robots have depended on teams of human data collectors and operators who guide them through various tasks—from squatting to washing dishes—using advanced technology like virtual reality headsets and motion-capture suits. However, Børnich highlighted that the new "world model" would enable Neo to autonomously learn from its own video footage, mimicking the guidance that a human operator would typically provide. This innovative approach not only enhances Neo's ability to generalize its learning but also significantly scales the robot's intelligence based on the number of units deployed rather than the number of operators. Børnich emphasized that this new model could potentially streamline the training process, making it less dependent on human intervention. The implications are vast. As the demand for robot training increases—whether for humanoids or large language models like ChatGPT—1X's strategy reflects a broader trend in the robotics field. Companies are increasingly turning towards autonomous data collection methods. A spokesperson from 1X mentioned that the new model greatly diminishes the need for teleoperation, suggesting that robots will soon gather most of their training data independently. The role of robot operators has been described as both physically taxing and monotonous, with reports of injuries among those working on projects like Tesla's Optimus. Positions typically offer starting wages of around $25 per hour. However, with 1X's new approach, the future of robot training may look less labor-intensive for humans. In October, Neo gained considerable attention with a viral demonstration showcasing its capabilities in chores like vacuuming and laundry folding. Priced at $20,000 as a one-time purchase or $500 per month through a subscription, Neo is scheduled for release later this year, although early users may still see some tasks requiring human oversight. Looking ahead, Børnich expressed confidence that significant advancements in the world model would allow Neo to operate autonomously in most scenarios by the end of this year. He projected that by 2026, the company could deliver fully autonomous robots that require no human interaction beyond the initial setup. With plans to produce over 10,000 units this year, 1X is already witnessing a surge in demand, having sold out shortly after the demo. As the robotics landscape evolves, 1X is poised to redefine the training paradigm, making it an exciting time for the industry and consumers alike.