Encord Explores Brain Wave Data to Overcome Robotics Training Challenges
Encord, a data tooling company based in San Leandro, California, is experimenting with brain wave measurements as part of its effort to address the scarcity of real-world physical training data for robotics. In collaboration with Zander Labs, Encord uses headsets equipped with sensors to track pilots' neural activity while they perform tasks like disassembling a Jenga tower. The goal is to determine whether incorporating mental states—such as intent and surprise—into datasets can improve the performance of robotic models. Encord's approach reflects a growing trend among robotics firms that are struggling to generate sufficient high-quality training data for physical AI systems, especially in areas like warehouse automation and humanoid robots. According to Encord’s head of robot learning, Vineeth Velmurugan, current datasets fall short by an order of magnitude compared to the scale needed—potentially five times larger than YouTube's video corpus—to enable breakthroughs similar to those seen with large language models. The company is also leveraging egocentric data from human workers and remote robotic operations. These efforts aim not only at managing existing training data but actively creating new datasets tailored for specific tasks, such as pouring coffee or stacking objects.
Topics: AI Data Training Neuroscience Robotics