Encord explores brain wave data to enhance physical AI training methods

Encord is conducting trials in a San Leandro warehouse, utilizing a brain wave headset developed by Zander Labs to collect data while operators perform tasks like disassembling a block tower.

The goal is to create a brain wave-tagged data set that could improve the performance of robotic models. Encord's head of robot learning, Vineeth Velmurugan, emphasizes the need for extensive real-world training data, which is currently lacking.

Robotics companies are increasingly relying on 'egocentric' video and remote operation data, with Encord experimenting with new data modalities, including brain wave measurements and muscle electrical signals.

Encord's data sets are meticulously annotated to enhance understanding for large language models, although generating this physical training data is significantly more expensive than collecting textual data from the internet.

The company aims to identify effective data techniques across the industry, leveraging insights from various robotics firms to improve physical AI models. Encord's workforce consists of skilled pilots who contribute to developing these innovative training methods.