Archive — US — techcrunch
Are brain waves the next unlock for physical AI?
Encord, a startup in San Leandro, California, is exploring brain wave technology to enhance AI training data for robotics. By measuring brain activity of trainers during tasks, they aim to create richer datasets for improving model performance, tackling the data scarcity challenge in physical AI. Their trials with Zander Labs highlight the innovative intersection of neuroscience and robotics.
Key points
- Encord is located in San Leandro, California, focusing on creating training data for AI models.
- Andrew Ceja, a pilot at Encord, uses a brain wave headset from Zander Labs to improve robotics training data.
- Current methods to gather training data include collecting egocentric video and remote robot operation data.
- Encord aims to build brain wave-tagged datasets to enhance performance of robotic models during task execution.
- The data required for training AI in robotics is realizing as scarce, necessitating generative approaches.
- Velmurugan notes that true advancements in training will require datasets significantly larger than existing online video collections.
Why it matters
- This represents a significant trend in AI development, focusing on the integration of neuroscience to enhance training datasets for more complex humanoid robotics.
- Similar to digital training for chatbots, physical AI challenges highlight the transition from data collection to data generation as a business necessity.
- The emergence of startups like Encord suggests a growing market for specialized data solutions in the robotics sector, reflecting larger trends in automation and machine learning.
Encord's pioneering approach to integrating brain wave analysis in AI training signifies a crucial evolution in addressing data constraints for robotics, potentially reshaping the future landscape of physical AI.