Tech Firms Eager to Capture Your Chore Time on Camera
This week, an AI training startup named Shift announced it would offer free home cleaning services to New Yorkers. They have plans to branch out to other cities, including London, and as I look around my apartment, I can see the appeal.
But, of course, there’s a catch.
In return for the cleaning, Shift wants video footage of its cleaners in action: scrubbing dishes, wiping down counters, dusting surfaces, and mopping floors. They’re after it all—capturing the mundane household tasks we’d gladly outsource if we could, and that robotics companies are racing to automate so they can sell us machines to handle them.
However, this is no easy feat. Unlike chatbots and image generators that have surged in popularity, robots must navigate the complexities of the physical world. This involves understanding space, motion, force, friction, and the various shapes and materials we encounter, along with the tricky lighting conditions that humans intuitively manage. It’s why tasks that seem simple to us, like folding laundry, picking up an apple, or pouring a glass of water, have proven to be incredibly challenging for roboticists to program.
To teach machines these skills, a vast amount of data is required. While text, images, and videos can be easily harvested from the internet, the physical world presents a tougher challenge. Gathering high-quality data without proper compensation is a significant hurdle for companies developing physical AI. This creates a lucrative opportunity, prompting companies like Shift to think outside the box.
They’re not alone in this endeavor. In India, reports have surfaced that home services platform Pronto has been using clients’ homes as a source for AI training footage related to chores like cooking, cleaning, and laundry. Pronto claims it only records footage if customers opt in, though it’s unclear what customers receive in return, aside from a copy of the footage. This practice has sparked backlash, with competing startups asserting they have never recorded inside homes for AI training and have no intention of doing so.
Other startups are focused on scaling data collection efforts. For instance, Silicon Valley-based Human Archive aims to collaborate with companies like Pronto, employing gig workers to record their activities using unassuming camera caps. These hats capture footage from the wearer’s perspective, providing the “egocentric” data that robotics companies need to teach machines how people navigate physical spaces. Meanwhile, Shift also engages consumers directly, claiming to have compensated tens of thousands of individuals across 15 countries to document their activities through its app.
Some companies are bypassing useful work altogether. Instead, they pay workers to repeat the same physical tasks while cameras and sensors capture every movement. These staged data farms are designed to transform repetitive physical activities—like folding towels, picking up cups, and carrying boxes—into valuable AI training material, justifying the compensation for those creating it.
Additionally, some data is generated by robots already operating in the field. Despite the buzz around automation, true autonomy is still a distant goal, which is why this data is so crucial. Companies are eager to release products, often using data from customers’ homes to refine their offerings. Many rely on remote workers to assist when robots inevitably encounter obstacles, and they utilize that data as well.
The concept of trading data for value isn’t new. Companies have long offered discounts, convenience, and free services in exchange for access to personal data, from loyalty programs and cookies to dashcams and insurance apps that monitor driving behavior, not to mention those intrusive smart TVs that constantly display ads.
What’s different now is the type of data companies are willing to pay for. For the time being, this might mean allowing a human to clean your home while wearing a stylish hat for free, paving the way for a future where a company can sell you a robot to do the job instead.
