Ken Goldberg has spent much of his life thinking about hands.
A human hand reaches for a glass and makes countless decisions without announcing them. Distance. Shape. Weight. Pressure. A robot must be taught each part. Turn the glass, change the light, add moisture, and the problem begins again.
Ken has built a career inside that uncertainty.
He holds the William S. Floyd Jr. Distinguished Chair in Engineering at UC Berkeley, with appointments in industrial engineering, operations research, electrical engineering, computer science, and art practice. Since joining Berkeley in 1995, he has directed AUTOLab, where researchers study robotic grasping, automation, and robot assisted surgery. He has published more than 300 peer reviewed papers and holds ten United States patents.
In 2026, Ken was elected to the National Academy of Engineering for his contributions to robotics and automation in industrial parts handling. His work grows from a stubborn question. How can a machine reliably pick up an object when it does not understand what it is touching?
Inside the lab, progress looks like patience. A gripper closes too early. A camera misreads the scene. Cloth slips. Someone adjusts the code. The arm resets and tries again.
Ken’s early research used geometry and mathematics to help robots orient objects with limited sensing. In 1994, he and his collaborators created the first robot people could operate through the internet. That work led to the Telegarden, where visitors directed a robot in Austria to plant seeds and water a living garden. It remained online for nine years.
For Ken, art is another way to test an idea. His installations have appeared in more than seventy exhibitions, including the Whitney Biennial. He founded Berkeley’s Art, Technology, and Culture Colloquium in 1997, creating a place where artists and engineers could press against each other’s assumptions. His later AlphaGarden project asked if artificial intelligence could learn to tend a crowded garden of different plants.
A garden is hard for a machine. Leaves overlap. Plants lean toward light. Soil dries unevenly. Nothing stays where it was.
The robots in Ken’s lab face the same disorder. They learn to sort packages, grasp unfamiliar objects, fold cloth, and assist with delicate surgical movements. His Dexterity Network research uses simulated objects and grasps to train robots to handle things they have never seen. That work helped lead to Ambi Robotics, which he cofounded with former students.
Ken remains skeptical of easy claims about humanoid machines. The physical world resists shortcuts. A robot must learn how weight shifts, how cloth collapses, and how one small error changes everything that follows.
During our shoot, Ken stood inside a cluster of robotic arms, his hands resting on two of them. The scene was playful and strange. The machines looked powerful, yet they remained still, waiting for human thought before making their next move.
The future in Ken’s lab felt close and unfinished. Aluminum frames. Rubber grippers. Camera lenses. A robotic hand reaches for an object and misses by millimeters. The motors hum as the system resets.
Then they try again.
