In this episode of Infinite Tech, Preston Pysh interviews Ken Goldberg, a robotics researcher, about the current state of AI and robotics. Goldberg addresses the inflated expectations surrounding AI's capabilities in robotics, particularly in manipulation tasks like tying shoelaces or buttoning shirts, which remain challenging due to the complexities of sensing and nuanced interactions with deformable objects. He highlights the progress in mobility with quadrupeds and drones but emphasizes the significant data gap between language models and robot manipulation, quantifying it as a "robot data gap" of 100,000 years. Goldberg also discusses his company, Ambi Robotics, and its focus on solving practical problems in logistics using AI-driven robots with simple grippers. He touches on the importance of good old-fashioned engineering alongside AI and shares insights on AI creativity and the potential for AI to learn from embodied experiences in physical forms.
Sign in to continue reading, translating and more.
Continue
