The Quest for Dexterity: Proception Raises $11M to Solve Robotic Hands
Startups like Proception are tackling the 'hand problem' in robotics by using novel data collection methods. Improving dexterity is seen as the final frontier for general-purpose robotic agents.
Dexterity remains the "holy grail" of robotics. While robots can now navigate terrain and identify objects with high accuracy, the subtle, complex movements of the human hand remain incredibly difficult to replicate. Proception, a robotics startup that recently settled a trade secret suit with Tesla, has announced an $11M raise to double down on solving this exact challenge.
Proception’s approach centers on a unique hardware-software pipeline for training data collection. Most robotic hands struggle with "contact-rich" tasks—actions like threading a needle or picking up a delicate fruit—where force feedback and tactile sensing are just as important as visual input. By using specialized teleoperation rigs, Proception is building a massive library of human-led tactile data to train their AI models.
The commercial implications are vast. A robot with human-level dexterity could revolutionize electronics assembly, home care, and laboratory work—fields that have so far resisted automation due to the fragility and variety of the objects involved. As Proception scales its data collection, the goal is to create a "foundation model" for manipulation that can be ported to any robotic limb, finally giving robots the hand-eye coordination they need to be truly useful.
Source: TechCrunch
