by Andrew Moseman
Autonomous robots, particularly the humanoid ones that could interact with us, go places too dangerous for humans, and be our personal assistants around the house, represent a next phase in the AI revolution. Technology companies big and small are investing heavily in robotics and joining the race to build intelligent mechanical helpers.
But fully realized humanoid robots represent a series of grand challenges on a much different scale than, say, AI chatbots, notes Aaron Ames, Caltech’s Bren Professor of Mechanical and Civil Engineering, Control and Dynamical Systems, and Aerospace; and the Booth-Kresa Leadership Chair and director of the Center for Autonomous Systems and Technologies (CAST). In addition to linguistic and conversational intelligence, they must also possess physical intelligence: the ability to navigate safely, locomote dynamically, and manipulate objects effectively. Ames appeared on a recent episode of PBS NewsHour’s Horizons to speak with host William Brangham on the opportunities and challenges of bringing humanoid robots into our homes.
New videos and demonstrations, including those that come out of the Ames lab, show robots that can walk, dance, balance, and sprint. However, Ames cautions against assuming that impressive demonstrations translate into practical everyday usefulness. Those robots, after all, are not navigating their environment with intelligence. They’re simply repeating what they’re taught to do.
“We’ve gotten really good at playing back motions on robots. We show the robot how to dance, how to pick up a raspberry. And they, after multiple tries, can do that with some success rate that’s actually not that great yet—50 percent of the time, or 70 percent, or even 80 percent. What do we need for reliable deployment? Well, we need to have a 99.9 percent success rate in labs, because when you go to the field, everything’s different”
Indeed, a robot that successfully performs a task in a laboratory may struggle to repeat it in a real home, where lighting changes, objects vary, floors are uneven, and people move unpredictably. Household assistance, from folding laundry to helping older adults, requires far more than the ability to execute a single task. It demands adaptability, dexterity, and dependable performance across countless situations. These capabilities require advances in robotics, control systems, and real-world learning.
Safety is another essential consideration. Robots operating alongside people must respond appropriately to unexpected obstacles and changing environments. Before humanoids become trusted household assistants, researchers must address these technical challenges, while considering the practical implications of technology companies deploying them in everyday spaces, Ames says.
“I love seeing robots in the home. I show off humanoids all the time to kids in my lab and see their fascination. [But I have mixed feelings:] I’m deeply passionate about this idea that robots can be part of our lives, and at the same time, my fear is that that excitement will take over rationality. Coupled with massive market pressures to deliver these things and deploy these things, they’re going to put it in places where they aren’t ready”