Key Takeaways
- Tau Robotics now offers a paid, invite‑only home‑cleaning service in San Francisco for $30 per hour, using humanoid robots named Chelsea, Elon, etc.
- The robots are currently supervised remotely by human technicians who can intervene via VR goggles and place the machines in a safety‑mode “lie down” if needed.
- Tau plans to scale to 1,000 cleanings a week by 2027 and expand nationally and internationally by 2028, aiming to phase out human supervisors once the bots achieve higher autonomy around 2028‑2029.
- Experts warn that the robots remain clumsy, pose physical safety risks (e.g., falling, pinching), and raise privacy concerns because their onboard cameras record video inside homes.
- Market analysts forecast rapid growth in the humanoid‑robot sector—potentially reaching a $200 billion valuation by 2035—but agree that widespread, reliable home use is still at least five to ten years away.
A New Pilot Program Brings Humanoid Maids to San Francisco Homes
Tau Robotics, an engineering firm that builds AI‑driven robots, announced last month that it is letting San Francisco residents hire a humanoid robot to clean their houses for $30 an hour. The service is part of an invite‑only pilot program open to 1,000 households. The robots, dressed in black Tau Robotics uniforms and given names such as Chelsea and Elon, perform tasks like vacuuming, wiping kitchen counters, and taking out the trash.
“Initially, they will be trained to imitate human performance, and then they will be trained to exceed human performance,” said Alexander Koch, Tau’s co‑founder and CEO, in an interview with CBS News.
Each unit costs roughly $50,000 to manufacture, and the company relies on artificial intelligence to teach the bots how to carry out chores.
Training, Scaling, and the Road to Full Autonomy
While the robots are currently learning by mimicking human actions, Tau’s long‑term goal is for them to exceed human efficiency. Koch revealed that the firm intends to scale up to 1,000 cleanings a week in San Francisco by 2027, then expand to other U.S. cities and internationally by 2028.
The roadmap includes a gradual reduction in human oversight: “We’re still pushing the limits of what it can do, so we are not entirely sure ourselves where the limits are,” Koch admitted. He estimates that by 2028‑2029 the bots will reach a level of self‑sufficiency that allows Tau to phase out the remote technicians who presently supervise them.
Human Supervision and Safety Mechanisms
At present, the robots do not operate completely independently. They are controlled remotely by human technicians who wear VR goggles to see through the robots’ “eyes” and can step in if a bot gets stuck. If something goes wrong, technicians can put the device into a safety mode that essentially commands it to lie down on the floor.
Koch acknowledged the current limits: “It doesn’t know how to walk upstairs, for example. It doesn’t know where to vacuum. That’s directed by mostly humans at the moment.” This reliance on human oversight is intended to keep both the robots and the household occupants safe while the AI continues to learn.
Expert Warnings About Physical Safety
Ken Goldberg, an engineering professor at the University of California, Berkeley who studies robotics and automation, cautioned that the humanoids are still “relatively clumsy” despite their sleek appearance. He noted that the robots are “actually quite heavy” and can fall over, potentially thrashing and striking anyone nearby.
“You wouldn’t want this thing anywhere near a child or a baby,” Goldberg said, emphasizing the risk of injury when the bots attempt to regain balance.
Tau’s technicians monitor proximity to prevent humans and robots from getting too close, but Goldberg’s remarks highlight that the technology is not yet ready for unsupervised interaction in homes with vulnerable occupants.
Privacy and Security Concerns
Beyond physical hazards, experts have raised questions about security and privacy. Tau’s robots are equipped with two cameras that serve as their “eyes,” recording video used to train the company’s AI. Goldberg warned that these cameras could inadvertently capture sensitive personal items—such as receipts, bills, or documents—inside a residence, potentially violating occupants’ privacy.
When asked about safeguards, Koch responded that Tau retains “full control” over the data it collects: “The camera system, the networking and the computers are all our own.” He asserted that the firm can manage how the footage is stored and used, but the assurance does not eliminate the underlying concern that a device constantly watching inside a home could be misused or hacked.
Market Outlook and Industry Growth Projections
The humanoid‑robot sector is attracting significant investment, driven by advances in AI. Barclays Investment Bank estimates the current market is valued at $2–3 billion but could swell to $200 billion by 2035. Morgan Stanley Research predicts that the global humanoid‑robot population may reach nearly 1 billion by 2050, though it stresses that prices must fall substantially and societal acceptance must increase before these machines become commonplace in households.
In 2025, Barclays counted roughly 15,000 humanoid robots deployed worldwide, with about 90 % employed in industrial settings such as manufacturing and warehousing. Analyst Zornitsa Todorova of Barclays described the adoption trajectory as “gradual and incremental,” forecasting that meaningful adoption in home services will take at least five years but that the economic opportunity for humanoids performing service‑oriented tasks is significant.
Goldberg, however, remains more skeptical, estimating that it will take another decade for the science to mature enough for robots to reliably handle at‑home chores like cleaning.
“I’m not saying we will never have them,” he said. “But we’re not close to having humanoid butlers. The science isn’t there yet.”
What Lies Ahead for Home‑Based Humanoid Robots?
Tau Robotics’ pilot offers a tangible glimpse into a future where robots share domestic labor, but the journey is fraught with technical, safety, and ethical hurdles. While the company’s ambitious timelines aim to reduce human supervision by the end of the decade, experts like Goldberg urge caution, pointing out that the robots still struggle with basic navigation and pose potential physical risks. Privacy advocates remind us that ever‑watchful cameras inside our homes demand robust data‑protection policies.
If the industry can overcome these challenges—improving agility, lowering costs, earning public trust, and securing data—humanoid helpers may eventually become as ordinary as vacuum cleaners. For now, however, the technology remains a promising but nascent experiment, with humans still very much in the loop, both as supervisors and as the ultimate judges of whether a robot truly belongs in the home.
https://www.cbsnews.com/news/tau-robotics-humanoid-ai-cleaning-robots-san-francisco/