How likely are we to see humanoid robots capable of performing surgery, replacing factory workers or serving as in-home butlers within a few short years? Leading roboticist Ken Goldberg, a CITRIS researcher and UC Berkeley professor of industrial engineering and operations research, weighs in on this question in two new papers published in Science Robotics.
Artificial intelligence (AI) chatbots, powered by large language models (LLMs) trained on vast sets of text data found on the internet, have rapidly gained traction as a versatile, albeit still improving, tool in professional and everyday life. Goldberg argues that robots face a “100,000-year data gap” to achieve similar proficiency in real-world skills. One challenge lies in dexterity, or the ability to manipulate objects, and researchers anticipate that complex, three-dimensional tasks will require much more data than the amount used by an AI chatbot to gain language fluency.
“AI systems can play games like chess and Go better than humans, so it’s understandable that people think, ‘Well, why can’t robots just pick up a glass?’” said Goldberg. “But the fact is that picking up a glass requires that you have a very good perception of where the glass is in space, move your fingertips to that exact location and close your fingertips appropriately around the object. It turns out that’s still extremely difficult.”