RLWRLD Captures Human Skills to Train Robot AI, Targeting 2028 Industrial Deployment
Updated
Updated · The Associated Press · May 12
RLWRLD Captures Human Skills to Train Robot AI, Targeting 2028 Industrial Deployment
11 articles · Updated · The Associated Press · May 12
RLWRLD is recording skilled workers at Lotte Hotel, CJ warehouses and Lawson stores to build a data library that trains robots to handle delicate real-world tasks.
The South Korean startup says hand dexterity is the key bottleneck, so it converts body-camera footage into machine-readable motion data and adds VR, glove and force-tracking sessions to refine robot behavior.
RLWRLD unveiled its robotics foundation model last week and expects industrial AI robots to be deployed at scale around 2028; Lotte hopes humanoids can take on cleaning and other back-of-house work by 2029.
South Korea is backing that push with a $33 million project to digitize master technicians' know-how as it tries to raise productivity and offset an aging, shrinking workforce.
Labor groups warn widespread robot adoption could cut jobs and weaken the pipeline for skilled workers, even as manufacturers such as Hyundai and Samsung plan broader humanoid and AI-factory rollouts.
Can AI robots save Korea's economy without creating an energy and water crisis?
If AI can learn in simulations, is copying human workers already an obsolete strategy?
Who owns a worker's skill once it's digitized to teach a robot their job?
RLWRLD’s $26M Bet on Physical AI: Inside the Race to Industrial-Scale Dexterous Robotics by 2028
Overview
RLWRLD is rapidly advancing toward its 2028 goal of large-scale industrial deployment of physical AI robots, positioning itself as a major player in robotics. Backed by a $26 million funding round, the company is leveraging the strong manufacturing and logistics infrastructure of South Korea and Japan to build a dominant regional presence before expanding into North America in late 2026. Central to its strategy is the upcoming launch of its core robotics intelligence layer, RLDX-1, which uses scalable data collection and a multi-stream transformer architecture to overcome current model limitations, driving innovation in industrial automation.