$30/hr. Video shows robot is very capable. Presumably arrives by Uber, walks in, does its work, and leaves by Uber to next job.
You wonder how many hours needed to breakeven. Who owns it. Looks like many are willing to give it a try.
Who wooda thunk house cleaners would be first to be replaced. Motels must want one (or more).
I am guessing that most maids don’t make $30 an hour - so it likely never breaks even based on that factor. This probably works in SF because it is so expensive there, and maid work likely demands a substantial premium.
Gemini suggests the median hourly rate is $20-$24. I doubt most get any significant benefits beyond that.
I pay $145 for a crew of three to show up and clean for 2 hrs. Very close to 25/hr.
Much depends on how much cleaning a robot can do in an hour. Also how much they cost. Reliability. Battery life. Etc etc. They could be far more available.
Video makes them look very capable. You wonder if that is AI or real.
Sooooo, still not cost effective.
There is no way that robot can clean quicker than a human. I watched how slow it moved. There is no comparison. Additionally, we did not see a demonstration of how it handles dirt it missed on first attempt. A human can very easily identify a counter or a floor that still has some dirt or grime that needs to be removed, sometimes with a bit of elbow grease (like dried spills on a kitchen counter). That robot does not appear to have the weight or the leverage that may be required for such. How does it even dust higher surfaces that it can’t see? Does it climb a ladder?
::chuckle:: Not my impression at all. Maybe for a small apartment that has been pre-cleaned but not for real life.
Show it cleaning the shower. Show it deciding that it needs to use more pressure to remove a stubborn spot. What does it do if it encounters a sheet of homework on the floor? Can it identify an important sheet of paper of a random piece of trash?
I think we will be watching to see how it does. They have made a beginning. And we assume working on improvements. Will take a while to reach the rest of the country. Meanwhile users will give their reports.
We pay $45/hr (each) for two people from a service, 3 hours once a month. $250 plus we give them $20 tip, and $100 at Christmas. They are fabulous and worth every penny.
I had to laugh at this robot demo. It couldn’t do what they do in 10 hours at the speed it was going. Ours tear apart the couch cushions, move furniture, do a complete clean of 3 1/2 bathrooms, kitchen, mop the floors, dust everything and more. They arrive with all their own equipment (does the robot do that?) and supplies, and make pleasant chatter if we happen to be home at the time.
Good luck replacing that at any price.
They’re not ever going to breakeven. That’s not the point.
All of these robots are teleoperated. There’s a human running the robot. That human alone is almost certainly going to be vastly more expensive, per hour, than a human housecleaner.
So why do this? Data. Data, data, data.
As we’ve discussed ad nauseum, the reason that humanoid robots aren’t anywhere near ready to enter useful work is the absence of data. LLM’s and chatbots and AI can do useful work writing things or coding software and making videos because those are entirely digitally native things to do. And humans have spent billions and billions of hours voluntarily creating massive publicly available libraries of examples of all those things, which we call the Internet. Billions and billions of texts. Billions and billions of captioned and labeled photographs. Billions and billions of videos. Etc.
No such repository of training data exists for real-world robots. There’s videos of humans moving things in the world, but those videos don’t have data explaining how much friction a surface provides, how much force was needed to hold or lift an object, etc.
So there are companies that are trying to create artisanal handcrafted datasets on which to train robots. The only way to get the data is to have humans just doing the tasks. You hire humans and have them wipe a counter or lifting plates or draw the curtains, over and over again. Even Tesla:
It’s expensive to do this, so it’s nice to have some of the cost defrayed a bit. Hence, you do the data-training in people’s houses and say it’s a housecleaning service. You get press, you get some portion of the cost covered, and you get variety in the cleaning scenarios your training data has.
But that’s what this is. Data capture, not an actual cleaning business. We’re many many years away from that, if ever we get there with humanoid robots.
The robots seem capable of house cleaning. Data collection makes possible automation. As with self driving cars accumulated data makes for better performance.
Not there yet, but a step in the right direction.
The robots can and will get better over time and possibly get cheaper. Your manual labor cannot get any more better. Just a thought…
Time is on the side of technology.
Well, humans teleoperating robots seem capable of house cleaning. I’m not sure that’s quite the same as the robots being capable of it. Given the paucity of real world data that exists (other than video, which can’t capture most of the information on forces that the robots need to learn), I suspect we’re a very long time away from the robots being able to do this.