This week, an AI training startup called Shift said it would clean New Yorkers’ homes for free. It also has plans to expand to other cities, including London, and it appeals to me as I look around my flat.
but there is a problem. There is always a catch.
In exchange for cleaning, Shift wants footage of its cleaners at work: scrubbing dishes, wiping counters, clearing tables, mopping floors. It wants everything. Video of all the boring household labor we’d happily outsource if we could — and robotics companies are racing to teach machines to do it so they can sell us something to do it for us.
It is more difficult than it seems. Unlike chatbots, image generators, and other AI tools that have exploded in recent years, robots have to deal with the physical world. That means space, motion, forces, friction, strange shapes and materials, strange lights, and everything else that humans – and other creatures – intuitively understand. This is why things that are usually easy for us, like folding clothes, picking up an apple, or pouring a glass of water, have proven so difficult for roboticists to code.
Teaching machines to do those things requires data. That’s a lot. Text, images and videos can be easily extracted from the Internet on an industrial scale. And they were often without compensation to the people who made them. It’s hard to scrape the material world, and it’s even harder to scrape quietly without paying for it. This means that access to high-quality data is a major hurdle for companies developing physical AI. It’s a lucrative opportunity, so companies like Shift are getting creative.
They are not alone. in india, recent reporting It turns out that home service platform Pronto is using customers’ homes as a source of AI training footage for tasks like cooking, cleaning and laundry. Proto says it only records footage if customers explicitly choose to do so — it’s not clear what customers get in return besides a copy of the footage — but the practice has still caused a wave of backlash in the market with rival startups. emphasis They have never recorded indoors to train AI and have no plans to do so.
Other startups are focused on trying to increase data collection. located in silicon valley human archiveFor example, it’s expected to partner with companies like Pronto and have gig workers record their activities using non-stylish camera caps. The hats collect footage from the wearer’s perspective, exactly the kind of “egocentric” or first-person data robotics companies need to teach machines how people navigate physical space. Meanwhile, Shift also taps consumers directly, and claims to have paid thousands of people in 15 countries to record their activities through its app.
Some companies are leaving useful work Completely. Instead, workers are paid to complete the same physical tasks over and over while cameras and sensors can capture every activity. As such Phased Data Farm Rote physical activity – folding towels, lifting cups, carrying boxes – is designed to be turned into AI training material to justify paying people to create it.
And some data is generated by robots already present in the world. Despite the hype, real automation is still a long way off – hence the need for all this data – but companies are eager to ship products anyway. They will use data from customers’ homes to improve the product. Many companies rely on remote workers when robots inevitably get stuck. They will also use that data.
Of course, trading data for something valuable is nothing new. Companies have been offering discounts, convenience and free services in exchange for access to your data for years, from loyalty cards and cookies to dashcams, insurance apps that monitor how people drive and that disgusting smart TV that’s always showing ads.
What’s new is what kind of data companies are willing to pay for. For now, that means letting a human wearing a fancy hat clean your house for free so that, eventually, a company can sell you a robot to do the job.
