"Teletubbies" robots provide door-to-door cleaning services at the rate of 200 RMB per hour, featuring pure artificial intelligence.
Humanoid robots are now available to help with household chores in San Francisco.
Not bad, it's another important step towards entering family scenarios, after all...
This guy can wash his hands all by himself!!!
It's absolutely amazing: squeezing hand sanitizer, turning off the tap... and finally it even knows to grab a cloth to wipe its hands.
It just looks a little weird, with a Wi-Fi router on its head, like a Teletubby, right?
But that doesn't stop it from doing chores. Its mopping demo is very interesting: while mopping, when it sees rubbish on the floor, it will bend down to pick it up, turn around and walk a few steps to throw it into the trash can.
Is this exactly what my mom calls "knowing what needs to be done without being told" (not really)?
But, here's the catch.
It wasn't until I scrolled to the comment section that I found out all the above actions were remotely controlled. Purely manual intelligence.
To be honest, my heart sank the second I read that line...
But after calming down and thinking about it, there are still parts worth praising.
All the videos above are played at 1x speed.
You know, for most household robot demos on the market, 2x speed is the baseline, and some even have to be played at 4x speed to be barely watchable.
At the current stage, it's still too difficult for humanoid robots to both perform tasks and maintain good viewing experience. Remote control is a rather clever workaround.
This is the debut of American robotics company Tau Robotics.
Yes, the τ of Embodied AI (not really).
The "Wi-Fi Router" That Does Housework
Tau Robotics is a US robotics startup founded in 2024, headquartered in the San Francisco Bay Area. It is a very early-stage small team with fewer than 10 employees.
Its development strategy is quite interesting.
It doesn't start with fully autonomous AI, but "cheats" through remote control to fill the technical gap. After realizing real family deployment, it can start the data flywheel.
In September 2024, it raised a Pre-seed round of about 3 million US dollars, and then remained silent for a long time. Now it finally presents its results —
Humanoid Robot Cleaning Service.
To be honest, at first glance I was wondering where the robot was.
After all, this guy is wearing a dark work uniform and looks just like a normal person when getting off the van —
Until I saw that Wi-Fi router on its head...
The founder also came out to explain it specifically, saying that it does look a bit funny, but there's no other way — this is the best position to place the antenna for the humanoid robot.
Bro, the problem is I didn't know it was remotely controlled at first. When I suddenly saw that antenna, I got goosebumps — has embodied AI developed to this level??
It's obvious that the official team is extremely satisfied with this demo of theirs. The video of the robot getting off the van loops infinitely in the background of their official website...
After entering the house, it starts working formally.
Tau has three "employees" under its belt.
The first one, Chelsea.
It is the guy that got off the van. It excels at working in kitchens and bathrooms: sinks, countertops, mirrors, tiles, floors...
The second one, Elon.
This name, little Elon Musk, right?
Elon is responsible for regular weekly visits. After working at your house several times, it can remember where all your items are placed, and put everything back in their original positions after cleaning up.
The third one, Tony.
Tony focuses on deep cleaning: skirting boards, tile gaps, window tracks, under the refrigerator, behind the washing machine, specializing in removing long-standing stubborn dirt.
But I really have to complain about one thing.
Tau, for your first public announcement, couldn't you just spend some money hiring a photographer to take proper official portraits?? I couldn't find where Tony was at all...
I won't go into details about the demos, since they are all remotely controlled, theoretically they can do any tasks that humans can do.
Wiping tables.
Mopping the floor.
Taking out the trash.
Finally, the question that everyone cares about most — what's the price?
It's acceptable, 30 US dollars per hour (about 200 RMB).
To be honest, this price is quite affordable in the US. For a human cleaner, you can easily pay 50 to 60 US dollars per hour.
Humanoid Robots Entering Family Scenarios
Last week I attended WAIC and talked with many friends working on embodied intelligence. A common conclusion is that it is extremely difficult for humanoid robots to enter ordinary households.
There are some attempts in China as well, but they are still in the initial stage. For example, the intelligent cleaning service launched by Ziliangshu in cooperation with 58.com some time ago is a very good case of robots entering family scenarios.
However, many netizens complained about it, saying that while the robot is working in front, a human cleaner still needs to stand by and supervise it. What's the point of that?
Alas, there's really no other way. Embodied AI faces an extremely difficult starting situation. Teams can only get robots into real scenarios first, otherwise they can't get any data at all.
But at present, the main focus in China is still prioritizing industrial scenarios, where the environment is relatively controllable, and it can also provide gradient optimization directions to build a closed data loop.
The logic in the US is different. The demand for manufacturing robots is not as strong as in China, which instead has spawned some companies that start from family scenarios. For example, 1X Neo, which became a little popular last year, also focused on the concept of entering households through teleoperation.
After all, the technical gap is really huge, and remote control is a rather smart approach. It is somewhat similar to the shadow mode in autonomous driving, which ensures work efficiency while collecting data at the same time.
But whether consumers are willing to pay 30 dollars an hour to hire a remote-controlled robot with built-in monitoring to work in their homes is another story...
Nine months after its release, 1X Neo has not yet achieved large-scale formal delivery, so it is hard to tell how the market will react.
So let's get back to the technology itself.
Some netizens are confused: Why do we have to make humanoid robots for household use?
The narrative that bipedal robots are more adaptable to unstandardized scenarios has been talked about to death. Even ordinary sweeping robots can move freely in most homes, aren't wheeled robots great enough?
I talked about this topic with the team of Sudu half a month ago. Their point of view is that the root cause lies in the data side —
If you ask a human operator to teleoperate a wheeled chassis to perform complex mobile manipulation tasks, the difficulty will be extremely high. For humanoid robots, operators can map their own body instincts directly to the robot, which greatly lowers the operation threshold.
From this perspective, the reason why Tau and 1X insist on making humanoid robots is largely for the convenience of remote operation.
It's a pretty first-principle-based choice.
Of course, there is another point that cannot be ignored.
Humanoid robots are absolutely top-tier in terms of providing emotional value.
A humanoid robot walking around your house, wiping tables, mopping floors and taking out the trash, while you sit on the sofa watching it... Well, paying 30 dollars to experience the life of an ancient Athenian citizen is not that bad (joking).
At present, Tau's "Teletubby" robot service adopts an invitation system. If you are in San Francisco, you can go to their official website to sign up for a try (joking).
"Athenian citizen" experience link: https://www.tau-robotics.com/
References:
[1]https://x.com/alexkoch_ai/status/2082135074615763251
This article is from the WeChat Official Account "QbitAI" (ID: QbitAI), author: Focus on cutting-edge technology, published with authorization from 36Kr.