Chelsea Finn explains data-driven robotics failures

youtube.com8 min read

In an interview, researcher Chelsea Finn discusses the extreme challenges of robotics and the data strategies needed to achieve physical intelligence. She explains how months of zero-percent success rates in tasks like laundry folding ultimately revealed the variability and sensor-to-action translation hurdles that generalized machine learning models must overcome. Open article →