Robot-training startup will send humans wearing cameras to clean your home

The latest twist in paying humans to wear head cameras for robot training data.

See full article...

 

[Sarty]

All other issues aside, you have to wonder about how representative the training data would be. The people who would sign up for this would seem to be extremely self-selecting.

Big bold nope! from me!

edit: not just "self-selecting" in terms of willingness to opt into the panopticon, but, like, how many New Yorkers have stairs internal to their dwelling unit?

 

[Madestjohn]

Something something .. if the product is free .. something something

 

[itanod]

Hell no.

 

[Shanrak]

I actually applaud the attempted application here, cleaning and other manual tasks like this is what we should be trying to automate away.

I still wouldn't sign up for this due to privacy concerns, however.

 

[CrystalCowboy]

Meanwhile, the company has spread Craigslist postings targeting residents of other US cities such as Boston...

Appropriate verb choice.

 

[bignick243]

"From a privacy standpoint, the Shift app website’s FAQ states that “names, faces or other personal information is automatically anonymized, with any sensitive details blurred before it’s ever used…. We blur all personally identifiable information from screens and ID cards, to pieces of paper and cell phones to help protect both you and your home.”


Riiiiiiiiight........ I'm sure they would provide proof of them doing that or you could just take their word for it. Unless they actually showed you how would you even know.

Also, wasn't there a story a while ago about a guy with a knock off roomba and he figured out it was mapping his house and sending the information back to the mothership?

 

[icwhatudidthere]

Why not advertise to AirBnB hosts?

 

[Missing_Linc]

Why not advertise to AirBnB hosts?

AirBnB hosts would demand to be paid for the free cleaning.

 

[KrookedRooster]

If it is so trustworthy I would want the CEOs and all other managers to make use of the "Free Service" and prove that their task data is included in the dataset.

I still wouldn't use it but just ONCE I want to see the product actually being used by those selling us on how "life changing" and "essential" it is.

 

[raxadian]

A reminder that people actually have to pay Amazon to spy on them with their spyware Ring smart doorbells. At least this is temporary free.

 

[EnPeaSea]

I've seen this one! First, just don't, but if you do, please, be kind to your bots!

 

[Sarty]

I still wouldn't use it but just ONCE I want to see the product actually being used by those selling us on how "life changing" and "essential" it is.

Some credit where credit is due, I think--to the extent you trust the company's good faith * and ability to actually execute their promises *, this seems pretty straightforward and up-front, in a literal quid-pro-quo this-for-that sense. You get free cleaning, and they get to film your junk and use the recordings subject to certain stipulations. That's not being sold as life-changing or essential, just a question of whether you think that's a good deal to you.

 

[xrmb]

I hope anyone using AI to do/augment their job knows they are training their own AI replacement. I have quite a few tricks to be productive and fast, now Claude Code is watching me use them all day long to help him out of the problems he caused... and I have seen him learn from it locally... that all will bubble up over time. Sure it will make everyone's experience better... but also maybe unemployed?

 

[James Shewey]

 

[randomcat]

I would already find it really weird to have strangers in my house cleaning it. Adding the idea that they will record in-depth footage to be endlessly analyzed and no doubt carelessly stored makes it seem like rather than appealing to me to use their service, this company is intentionally trying to scare me.

 

[Madestjohn]

I would already find it really weird to have strangers in my house cleaning it. Adding the idea that they will record in-depth footage to be endlessly analyzed and no doubt carelessly stored makes it seem like rather than appealing to me to use their service, this company is intentionally trying to scare me.

That is the established standard marketing for AI

Tell people it will replace them, take their jobs, change every single aspect of their lives, and maybe destroy the world,
Then let the FOMO drive adoption.

 

[Xepherys]

If it is so trustworthy I would want the CEOs and all other managers to make use of the "Free Service" and prove that their task data is included in the dataset.

I still wouldn't use it but just ONCE I want to see the product actually being used by those selling us on how "life changing" and "essential" it is.

I would be shocked if they didn't already have data for their homes in this. Having worked for several tech start-ups and small tech companies over the decades, usually the folk with the idea use their own resources first to ensure the concept is even tenable before committing money to expand datasets.

 

[angleiron]

End of the day, this is why humanoid robots actually make sense. Training is straightforward compared to weird forms. Just record people doing things.

Right, because people never, ever do things that are unpredictable, stupid, sloppy, clumsy, criminal, etc.

 

[Person_Man]

I'd totally sign up for this if I was in the area, of course I have nothing too valuable and it's going to take way more than 2 hours to clean the dump.

 

[Alfonse]

End of the day, this is why humanoid robots actually make sense. Training is straightforward compared to weird forms. Just record people doing things.

Not really, no. Even for general cleaning, it's much easier to design a machine that can navigate and interact with the world that doesn't also have to deal with all of the thing a humanoid form needs to.

 

[bob_page_cousin]

Having footage inside people's homes huh?
Well, my three pals all like to wear a balaclava.
They are very, very interested in seeing this footage.
They just like your interior decor.
They wouldn't ever do anything to hurt you.
You can trust me on that.

 

[Mentil]

If they extend this business model to training sexbots the same way, I'd sign up...

 

[JoHBE]

$20 per hour to render yourself and other members of your species obsolete.

Personally, I would at least want to be made a millionaire.

 

[ampet]

This is literally what happened during the dotcom bubble. Companies offering services below cost in the expectation of capitalising later on.

 

[hildey328]

Eff a buncha that shizzz. Far, far worse than the Roomba sending interior maps back to HQ.

 

[42Kodiak42]

End of the day, this is why humanoid robots actually make sense. Training is straightforward compared to weird forms. Just record people doing things.

Say, have you seen what Boston Dynamics' humanoid robot looks like? Because it demonstrates a massive problem with "humanoid robots" so you can use training data from humans: Human muscles and robotic actuators work in completely different ways, resulting in wildly different weight distributions across the body, resulting in significant differences in how humanoid robots and humans move.

To use training data in the way you're suggesting, your humanoid robot designs would be required to have a similar weight distribution as an actual human because a top-heavy robot is just going to fall over if it tries to bend over while balancing itself with human training data.

Ironically, centaur shaped robots would have an easier time using training data from humans because the bottom half can keep the robot balanced regardless of what the upper torso is doing.

 

[Madestjohn]

$20 per hour to render yourself and other members of your species obsolete.

Personally, I would at least want to be made a millionaire.

My union sent me a email awhile ago announcing they had partnered with a AI company and were offering members an opportunity to have AI agents shadow their workflow in anticipation of the resulting productivity gains.

… this is IATSE, the main film production union.

cognitive dissonance has won

 

[Fred Duck]

I hope anyone using AI to do/augment their job knows they are training their own AI replacement.

Except in this case, it's everybody's replacement. Wonderbar.

A tech startup is offering New York City residents free home cleaning with a twist—it will send “professional cleaners” wearing cameras to record everything they do.

Every time I see "professional cleaner" in quotation marks, I assume they mean a Killer-7 type of "professional cleaner."

 

[Resistance]

End of the day, this is why humanoid robots actually make sense. Training is straightforward compared to weird forms. Just record people doing things.

I ask this of every humanoid robot evangelist none have presented a reasonable answer, most don't present one at all:

Name one task in the history of automation that would have been better done by humanoid robots.

Of course the response of the not completely irrational evangelists is to suggest that humanoid robots would be superior at tasks that we have not successfully automated.

When asked which tasks they have in mind, I rarely get anything remotely compelling, usually things like:
Food preparation (we have successfully automated this, microwave dinners have been a thing for decades, sliced bread has been a thing for just a hair under a hundred years. the counter argument is quality issues, but those aren't an artifact of automation). Would a humanoid robot kneading bread and cutting it into slices one at a time have been a better option?

Cleaning: we have actually done a lot of work to automate this and the work is ongoing, vacuum cleaners, dishwashers, washing machines, premixed cleaning chemicals. Self cleaning bathrooms have also been a thing for a very long time. Would a humanoid robot washing each dish one at a time with a sponge and running water have been a better solution than a human loading a dishwasher?

Construction: we have been automating this for a long time too, from prefabricated components like drywall, to power tools like nail guns. Would a humanoid robot that applies plaster to slats of wood have been a better option than a human screwing drywall into studs?

There will always be non humanoid systems that will be superior to humanoid robots. And adoption of humanoid robots will take an absurdly long amount of time, as evidenced by the vast amounts of human labour that has automated alternatives that are not universally adopted.

Sure a hypothetical humanoid robot that has the same abilities as a human doing the above tasks would be a neat tool to solve the "last mile" problem of automation, but to pretend that this is coming any time soon (humanoid robots have been in active development for decades, and will likely continue to be for quite a while longer before they become a better option than a human operating complex tools), and to pretend that it will be more affordable than a human right when that happens, is so absurd I can only laugh.

 

[C-Port]

I actually applaud the attempted application here, cleaning and other manual tasks like this is what we should be trying to automate away.

I still wouldn't sign up for this due to privacy concerns, however.

And they are trying to be open and honest about the data collection, instead of scraping the web and illegally downloaded books!

And compensating people for the training data, by presumably paying the people cleaning, and providing free cleaning to the people who sign up with clear sounds terms laid out.

Now sucks to train your own job replacement. I feel for the people wearing the cameras. I don't want to brush that aside.

That said, this sounds like the most ethically sourced training data I can imagine. Unless I'm missing something?

 

[jbjhillbilly]

Hard no.

 

[Madestjohn]

I ask this of every humanoid robot evangelist none have presented a reasonable answer, most don't present one at all:



Of course the response of the not completely irrational evangelists is to suggest that humanoid robots would be superior at tasks that we have not successfully automated.

When asked which tasks they have in mind, I rarely get anything remotely compelling, usually things like:
Food preparation (we have successfully automated this, microwave dinners have been a thing for decades, sliced bread has been a thing for just a hair under a hundred years. the counter argument is quality issues, but those aren't an artifact of automation). Would a humanoid robot kneading bread and cutting it into slices one at a time have been a better option?

Cleaning: we have actually done a lot of work to automate this and the work is ongoing, vacuum cleaners, dishwashers, washing machines, premixed cleaning chemicals. Self cleaning bathrooms have also been a thing for a very long time. Would a humanoid robot washing each dish one at a time with a sponge and running water have been a better solution than a human loading a dishwasher?

Construction: we have been automating this for a long time too, from prefabricated components like drywall, to power tools like nail guns. Would a humanoid robot that applies plaster to slats of wood have been a better option than a human screwing drywall into studs?

There will always be non humanoid systems that will be superior to humanoid robots. And adoption of humanoid robots will take an absurdly long amount of time, as evidenced by the vast amounts of human labour that has automated alternatives that are not universally adopted.

Sure a hypothetical humanoid robot that has the same abilities as a human doing the above tasks would be a neat tool to solve the "last mile" problem of automation, but to pretend that this is coming any time soon (humanoid robots have been in active development for decades, and will likely continue to be for quite a while longer before they become a better option than a human operating complex tools), and to pretend that it will be more affordable than a human right when that happens, is so absurd I can only laugh.

Humanoid robots are attractive to the C-suite for one reason - they believe they will somehow replace ‘humans’
Specifically employees
And as is typical of C-suite they have no idea what their employees actually do, that is as deep as the thinking goes.

 

[C-Port]

I ask this of every humanoid robot evangelist none have presented a reasonable answer, most don't present one at all:



Of course the response of the not completely irrational evangelists is to suggest that humanoid robots would be superior at tasks that we have not successfully automated.

When asked which tasks they have in mind, I rarely get anything remotely compelling, usually things like:
Food preparation (we have successfully automated this, microwave dinners have been a thing for decades, sliced bread has been a thing for just a hair under a hundred years. the counter argument is quality issues, but those aren't an artifact of automation). Would a humanoid robot kneading bread and cutting it into slices one at a time have been a better option?

Cleaning: we have actually done a lot of work to automate this and the work is ongoing, vacuum cleaners, dishwashers, washing machines, premixed cleaning chemicals. Self cleaning bathrooms have also been a thing for a very long time. Would a humanoid robot washing each dish one at a time with a sponge and running water have been a better solution than a human loading a dishwasher?

Construction: we have been automating this for a long time too, from prefabricated components like drywall, to power tools like nail guns. Would a humanoid robot that applies plaster to slats of wood have been a better option than a human screwing drywall into studs?

Sure a hypothetical humanoid robot that has the same abilities as a human doing the above tasks would be a neat tool to solve the "last mile" problem of automation, but to pretend that this is coming any time soon (humanoid robots have been in active development for decades, and will likely continue to be for quite a while longer before they become a better option than a human operating complex tools), and to pretend that it will be more affordable than a human right when that happens, is so absurd I can only laugh.

You say all reasonable things.

Am I correctly seeing the implication that we shouldn't try to do this?

It will take time. It will not be cheap for a long time. Will probably not work well for a long time.

Will never happen if we don't take steps towards it, however.

The world has a large problem with age distribution skewing old. We need a robot to be able to take the meal from a fridge, put it in a microwave, and bring it to the old person. Humanoid robot may or may not be a better solution than making a combo fridge/microwave/conveyor belt giant complex machine. Or multi-part compatible machine.

You mention the last mile equivalent problem here. What is your alternative?

 

[Argent Claim]

Sure a hypothetical humanoid robot that has the same abilities as a human doing the above tasks would be a neat tool to solve the "last mile" problem of automation, but to pretend that this is coming any time soon (humanoid robots have been in active development for decades, and will likely continue to be for quite a while longer before they become a better option than a human operating complex tools), and to pretend that it will be more affordable than a human right when that happens, is so absurd I can only laugh.

It is very telling, too, that people willingly ignore pioneering humanoid robots like ASIMO so they can maintain the illusion of the level of "progress" they desire by starting with more recent examples like Optimus instead. Then again, I often wonder how much of the humanoid robot hype is simply driven by Musk fans.

 

[kaibelf]

As someone who literally moved out of a large apartment this week and has yet to clean it before handing over the keys, I would be completely on board with this because all that they’re gonna get a view of is my filthy old carpets and dirty stove. It would save me a lot of trouble.

 

[Sarty]

I ask this of every humanoid robot evangelist none have presented a reasonable answer, most don't present one at all:

Name one task in the history of automation that would have been better done by humanoid robots.

I'm basically on your side on this one, but for the sake of fun on a Friday afternoon, I'll play. There are some circumstances where it would be nice to send in a worker that doesn't necessarily need to come back.

Right now, and maybe for a long long time, something like Chernobyl cleanup is technically inaccessible for rad-hardening reasons. But the old Coast Guard saying is you have to go out, you don't have to come back. It would be pretty neat if we could send robo-sailors in the very worst conditions and on the most doom-prone missions.

 

[Argent Claim]

The world has a large problem with age distribution skewing old. We need a robot to be able to take the meal from a fridge, put it in a microwave, and bring it to the old person.

1. Not all elderly people are confined to bed for significant periods of their life 2. How people age will continue to change with other technologies that may reduce the demand for the equivalent of live-in nurses and 3. You don't necessarily need a humanoid robot to do that. R2-D2 did just fine serving drinks on Jabba's barge.

Humanoid robot may or may not be a better solution than making a combo fridge/microwave/conveyor belt giant complex machine. Or multi-part compatible machine.

You're assuming a false premise here by constraining us to two options.

Complicated robots for such tasks can often be more trouble than they are worth. It's one thing to assume worming humanoid robots will be a thing, but it's another to further assume they will be able to do much of a variety of tasks at all for lack of training and payload capacity (especially when one considers how much heavy lifting one may end up doing with less mobile elderly), that they will be able to do such tasks reliably, and that they will be competitive with humans. It hardly matters if your robot costs more than a person whole requiring further money for maintenance, repairs and debugging.

More importantly, the focus on caring for an elderly population ignores an obvious solution: Don't let so much of your elderly population become so weak and dependent in the first place. We have the means to avoid a lot of that right now and, as I already mentioned, there's little reason to believe we're all fated to become bed-ridden wrecks waiting years for expiration.

 

[quamquam quid loquor]

I'm basically on your side on this one, but for the sake of fun on a Friday afternoon, I'll play. There are some circumstances where it would be nice to send in a worker that doesn't necessarily need to come back.

Right now, and maybe for a long long time, something like Chernobyl cleanup is technically inaccessible for rad-hardening reasons. But the old Coast Guard saying is you have to go out, you don't have to come back. It would be pretty neat if we could send robo-sailors in the very worst conditions and on the most doom-prone missions.

Humanoid robots would be really good at waging war, both on the logistics side and the warfighter side.