The article explores how companies like Encord and Zander Labs are leveraging brain wave data and advanced sensors to address the critical challenge of training physical AI systems.Encord, a data tooling company, is collaborating with Zander Labs to create brain wave-tagged datasets for robotics training.By measuring mental states like error and intent through EEG sensors, these datasets aim to improve the performance of humanoid and warehouse robots.

The piece highlights the scarcity of real-world physical training data as a major bottleneck for AI development, contrasting it with the abundance of text data used to train large language models.

Encord's approach combines egocentric video data from workers and robotic data collection, while experimenting with new modalities like muscle sensors and 3D hand tracking.

The article emphasizes the economic shift in AI development, where generating physical training data requires significant investment compared to text-based models.

Experts like Vineeth Velmurugan argue that the future of physical AI depends on creating massive, high-fidelity datasets, which Encord aims to provide through its global data collection network and partnerships with leading robotics firms.