While AI chatbots are making rapid progress, humanoid robots are not advancing nearly as quickly. This is the view of robotics expert Ken Goldberg from UC Berkeley. In two recent papers, he explains that robots face very different challenges – especially due to a huge “100,000-year data gap” and difficulties with everyday physical tasks. Language models learn from massive amounts of text data, but there simply isn’t enough data available for physical movements.
Goldberg sees robotics at a turning point: some focus on data-driven learning, while others rely on traditional engineering. He believes real progress will come from combining technology with practical, scalable data collection. For now, blue-collar jobs – such as those in trades and industrial work – are safe.