Humanoid Robots: Hype, Teleoperation, and Limited Data

Humanoid robots are back in the spotlight — and the coverage is equal parts fascination and skepticism. Robert Hart’s Stepback column traces the hype from Elon Musk’s Optimus demos (including a viral fall that revived questions about teleoperation) to a broad industry “gold rush”: Nvidia, Meta, SoftBank, Google, Amazon, Microsoft, Intel, Tesla, and startups such as Figure AI and 1X are all pushing humanoids. China has made embodied AI a national priority, spawning robot games, competitions, and a rush of low-cost models.

Despite flashy demos — videos of Neo doing chores, Ant Group’s R1 “cooking,” and humanoid fights — many demonstrations are staged, scripted, or remotely controlled. The piece emphasizes a core technical bottleneck: robots need real-world motion data for generalization, and that data is scarce. Companies are collecting it via teleoperation and in-home deployments (Tesla’s workers wearing sensors is one example), and some firms accept partial autonomy as a tradeoff to harvest training data.

Hardware costs are falling (Chinese models can be cheap; consumer humanoids range into the tens of thousands), but regulators and China’s planners warn of a potential bubble where investment outpaces real use cases. Hart’s takeaway: the tech may be inching closer, but until robots stop hiding behind promo videos and remote control, “eventually” remains an open question.

Source