Robots are good at brute force and repetition, but they remain clumsy at the kind of fine, fingertip-level manipulation that humans barely think about. A new collaboration aims to close some of that gap.
According to ABB, ABB Robotics and the prosthetics company PSYONIC are working together to advance robotic dexterity using human-generated data. The idea is to capture how people actually move and handle objects, then feed that information into robotic systems so they can perform more delicate, human-like tasks.
Forbes reports that PSYONIC is supplying this dexterity data to both NVIDIA and ABB. That framing positions PSYONIC as a data provider whose human-derived information helps train and improve the hands and control systems used by larger robotics and computing players.
The through-line across both sources is the same: human-generated data is the raw material for teaching machines to be more nimble. Rather than programming every grip and adjustment by hand, the approach leans on real human movement to give robots better instincts for touch and manipulation.
The sources here are brief, so the specifics of the arrangement—scale, timeline, and technical detail—are not spelled out. What is clear is the lineup: a prosthetics specialist (PSYONIC), an industrial robotics giant (ABB), and a chipmaker central to AI (NVIDIA), all converging on the same hard problem.
Why it matters: dexterity is one of the last big barriers keeping robots out of unstructured, real-world jobs, and using human data to crack it could speed the arrival of machines that can handle the messy, hands-on work people still do today.