← NewsAll
AI startup says it's relying less on humans to train its robot
Summary
1X announced a new "world model" it says will let its humanoid robot Neo learn from video the robot records, reducing reliance on human teleoperators; Neo is priced for consumers, expected to ship this year, and early units may still be teleoperated.
Content
1X, a robotics startup backed by OpenAI and Samsung, announced a new "world model" the company says will let its humanoid robot, Neo, learn from video it captures itself. The announcement frames the change as a shift away from large teams of human teleoperators who traditionally collect training data using VR headsets and motion-capture suits. Neo was shown in a demo performing household tasks and is offered at a one-time price or subscription; the company expects to ship units this year and said early adopters will still see teleoperation at times.
What we know:
- 1X says the new world model allows Neo to learn from video recorded by the robot rather than relying primarily on human operators.
- CEO Bernt Børnich said the approach should let intelligence scale with the number of deployed robots instead of the number of human data gatherers.
- Neo was demoed doing household chores and is priced at $20,000 as a one-off or $500 per month as a subscription, with shipments expected this year.
- 1X said it sold out shortly after the demo, expects to produce over 10,000 robots this year, and raised $100 million in 2024.
- The company and a spokesperson said the update "significantly reduces" teleoperation, but early units will still sometimes be piloted remotely and task execution may be imperfect.
- Other robotics efforts, including reported changes at Tesla, have also shifted toward using video and simulated "world models" for training.
Summary:
1X says the world model will change how data for humanoid robots is collected by shifting much training onto robots that record their own video. The company expects Neo to ship this year with some remote teleoperation for harder tasks and said it aims for a fully autonomous product in 2026. The degree and timing of reduced human involvement in customer homes will evolve as the company deploys and updates Neo.
