The robotics company behind the Neo humanoid robot, 1X, has unveiled a brand new AI model that it says understands the dynamics of the actual world and can assist bots learn recent information on their very own.
This physics-based model, called 1X World Model, uses a mix of video and prompts to present Neo robots recent abilities. The video allows Neo robots to learn recent tasks they weren’t previously trained on, in keeping with 1X.
This release comes as 1X is gearing as much as release its Neo humanoids into the house. The corporate opened up preorders for its humanoids in October with plans to ship the bots this 12 months. A 1X spokesperson declined to share a timeline of when these bots were shipping or share any information regarding what number of have been ordered beyond saying preorders exceeded expectations.
“After years of developing our world model and making Neo’s design as near human as possible, Neo can now learn from internet-scale video and apply that knowledge on to the physical world,” Bernt Børnich, founder and CEO of 1X said in an announcement. “With the power to rework any prompt into recent actions — even without prior examples — this marks the start line of Neo’s ability to show itself to master nearly anything you can think to ask.”
Saying that the bot can transform any prompt right into a recent motion is a lofty claim and never entirely accurate; you’ll be able to’t tell a Neo to drive a automobile and it’s going to suddenly know the way to parallel park, for example. But there’s some learning happening.
1X isn’t saying the world model allows today’s Neo bots to do a brand new task immediately from capturing video and being prompted, an organization spokesperson clarified. As an alternative, the bot takes video data linked to specific prompts after which sends that back into the world model. That model is then fed back into the network of bots to present them a greater understanding of the physical world and more know-how.
It also gives users insight into how Neo is pondering of behaving or reacting to a certain prompt. That form of behavioral information could help 1X train these models to a degree where robots will give you the option to react to a prompt of something they’ve never done before.
Techcrunch event
San Francisco
|
October 13-15, 2026