
Bionic Breakthrough: When Data From Prosthetic Arms Become Industrial Robot Training "Curriculum"
ARM - the chip design corporation that dominates 99% of the global smartphone chip market - is making strategic moves into a completely new field: Physical AI. By processing computing data directly on the terminal (On-device compute), ARM is collaborating with medical technology companies to turn rehabilitation devices into giant data collection stations for machine training.
The focus of this step is "Ability Hand" - a biological arm manufactured by Psyonic company (USA). This device weighs only 490 grams, lighter than a real human physiological arm, but is equipped with up to 6 micro motors and 30 multi-point tactile sensors. The sensing capabilities are so refined that the device can resolve forces from 10 grams to 1 kilogram, allowing users to hold a raspberry without crushing it, or immediately switch to lifting a kettlebell weighing 63 lbs (about 28.5 kg). Device assembly is also optimized in just 15 minutes thanks to flexible soft sockets technology from Koala company (UK), completely replacing expensive hard plastic molds.
However, the core value of the project does not stop at medical functions. The design of Ability Hand follows a modular mechanism: The same bionic hand that serves people with disabilities can be easily disassembled and assembled directly into an industrial robot arm, humanoid robot or dog robot. The developers believe that, to build a machine capable of performing sophisticated human tasks (such as washing dishes, folding clothes or sorting components on an assembly line), the most optimal solution is to use hardware that has been designed and tuned for the human body.
Every time the user makes a movement, internal sensors will record all pressure, speed and time of physical contact. This data is synchronized with the camera system to record real-life surroundings. All of this invaluable data is then fed directly into AI models to train machines to perform the same task. Using realistic physical data (Physical AI) helps the system eliminate the risk of "AI illusion" (Hallucination) - a vital factor to prevent the risk of software errors for medical surgical robots or heavy industrial machinery. It can be affirmed that the future generation of autonomous robots is being learned and perfected directly from the daily movements of current amputation patients.
Reference source: (2) How prosthetics tech is training robots | BBC News - YouTube | BBC News