The Real Challenges Slowing Down Human-Like Robots

At the 2026 Consumer Electronics Show (CES) in Las Vegas, technology correspondent Rowland Manthorpe explored how close robots truly are to matching human capabilities. During a staged fight with Unitree’s humanoid G1 robot, he found that while modern robots can balance and recover from impacts, they still lack human-like agility and adaptability. These limitations reveal deeper challenges in robotics: machines struggle not only to maintain balance but also to rethink tasks dynamically as humans do. The vision of a household robot assistant remains distant because replicating human movement, perception, and dexterity is extremely complex. One highlight of the expo was Dyna’s laundry-folding robot, capable of folding clothes with precision. However, adapting to new environments requires days of remote training, showing how fragile such progress remains. Similarly, companies showcased robotic hands trying to replicate human touch through tactile sensors and back-driveability—the capacity to yield under pressure rather than stiffen. WiRobotics demonstrated an actuator that allows a robot’s handshake to feel almost human, but perfecting the switch between softness and rigidity remains an unsolved problem. Despite these challenges, the integration of generative AI is transforming the field. Roboticists from Boston Dynamics and others believe this technology could help robots learn and adapt quicker than ever before, driving steady progress toward truly human-like machines. Nonetheless, experts agree that expectations for fully capable humanoid robots remain premature.