How UC Berkeley’s Ken Goldberg Is Teaching Robots to See, Learn, and Build

Ken Goldberg is a great resource to have on speed dial when those complex robotics questions emerge. As a longtime U.C. Berkeley professor and cofounder of Ambi Robotics, he’s familiar with the promise the category holds and the vast amount of work and data it will take to bridge that gap.

From Ambi Robotics’ data flywheel and the “robot data gap” to why personal use of humanoids may remain decades away, Goldberg and Brian Heater explore the hard truths behind dexterity, simulation, and learning at scale. Lock in for a synopsis of where robotics stands today and where it’s heading next.

Key Moments
00:00: Intro
01:10: What creativity means for machines
02:30: From Rosie the Robot to real robotics
05:00: Why humanoids are harder than they look
09:40: Why robots still struggle with grasping and dexterity
12:25: Why perfection is impossible in home robots
17:30: The humanoid timeline: hype vs. reality
20:00: The “robot data gap” and why AI can’t yet solve robotics
28:20: Four ways robots actually learn (Sim, YouTube, teleop, production)
34:30: Is Ambi Robotics a data company?
38:00: From picking to packing: solving the stacking problem
42:35: The long road to hardware-agnostic robots
45:40: Art, AI, and the creativity of machines

Use Ambi: https://www.ambirobotics.com/
Connect with Ken: https://www.linkedin.com/in/goldbergken/
Watch Ken’s 2025 GTC talk: https://www.nvidia.com/en-us/on-demand/session/gtc25-S74739/

Thoughts? Guest suggestions? Email us at podcast@automate.org.

You can find the transcript and more episodes of Automated at automate.org/podcast.

Unlock full access to Automated and explore everything automation. Subscribe today on YouTube, Apple Podcasts, and Spotify.

You can also find us on:
LinkedIn https://www.linkedin.com/showcase/automated-podcast-by-a3/
Instagram https://www.instagram.com/automatedpod/

Subscribe to the Automated Newsletter: https://www.automate.org/automation/newsletter-automation-roundup