Brain Wave Data Explored to Enhance Physical AI Training
Companies are testing brain wave sensors to improve robot training data. New technology combines visual data from robot tasks with brain activity measurements.
The frontier of physical Artificial Intelligence (AI) is exploring novel methods to augment training data for robots. One emerging approach involves measuring brain waves, which could provide deeper insights into robot performance and decision-making processes.
US-based Encord is currently trialing headsets equipped with brain wave sensors in collaboration with Germany's Zander Labs. Human operators performing physical tasks, such as disassembling a wooden block tower, wear these headsets. The technology captures not only the operator's visual field but also their brain wave activity during the task.
Zander Labs aims to utilize brain activity—deducing mental states like error, intent, and surprise—to create more valuable datasets for training AI models. Encord's objective is to verify if this data improves robotics model performance before scaling its implementation.
The scarcity of real-world physical training data has been identified as a significant bottleneck in robotics development. These new brain-wave-based methods aim to address this challenge by offering richer, more detailed information, potentially leading to more effective and responsive robots.