Training Even Brain Waves: Humanoid Competition Shifts into a Data War Beyond AI

AI data startup Encord is building a new robot-training dataset that combines egocentric video with brainwave and muscle-signal data from workers.

The company is testing whether these high-density sensor signals can improve learning for robots and humanoids in physical AI tasks.

The move reflects a broader bottleneck in humanoid AI: progress may depend less on model architecture than on the shortage of real-world data.

If successful, sensor-rich datasets could become critical infrastructure in the race to develop next-generation robots.