We Study Billionaires - The Investor’s Podcast Network TECH010: The Real Robotics Timeline w/ Ken Goldberg (Tech Podcast)
155 snips
Dec 24, 2025 
Ken Goldberg, a leading robotics researcher and founder of Ambi Robotics, dives into the challenges of robotics today. He addresses concerns about a disconnection between advanced AI language models and the physical manipulation capabilities of robots. Goldberg explains why tasks like tying shoelaces are still out of reach for robots, discusses the importance of tactile sensing in robotics, and highlights how simpler grippers can outperform complex robotic hands. He also sheds light on the vast data gap that hinders progress in this field and emphasizes the engineering behind real-world robotic applications.
AI Snips
Chapters
Transcript
Episode notes
Vision Can Compensate For Touch
- Surgeons perform complex manipulation largely using vision rather than tactile sensing.
- Visual-tactile inference suggests camera-based approaches could bypass some tactile hardware needs.
Add Sensors, Don’t Omit For Principle
- Use pragmatic sensors where they help performance rather than ideological purity.
- Goldberg endorses adding cameras to robot hands to improve manipulation feedback.
The Robot Data Gap Is Massive
- The robot data gap vs. language models is enormous — roughly a 100,000-year-equivalent of human reading time.
- Robotics lacks comparable large-scale paired sensor-to-action datasets to fuel current AI methods.