Backed by companies like Panasonic and Daiwa House, Laundroid had ambitious dreams to be the ultimate wardrobe organizer for the entire household. It had multiple cameras and robotic arms to scan a load of laundry, and used Wi-Fi to connect to a server that would analyze the clothing using AI to figure out the best way to fold it. A companion app was supposed to be able to track every piece of clothing that went through Laundroid, and categorize the clothes by household member. One load of laundry would take a couple hours to be folded, as each T-shirt took about five to ten minutes.
That’s how it was supposed to work in theory, anyway — when I tested it out at CES 2018 with my own T-shirt, the machine ate it up and Laundroid engineers had to work for about 15 minutes to pry it out. The explanation was that its cameras couldn’t recognize my black shirt, only the brightly colored demo shirts they’d prepared on hand.Dami Lee, “The company behind the $16,000 AI-powered laundry-folding robot has filed for bankruptcy” at The Verge
Good thing it wasn’t a name-and-number jersey for that evening’s game…
Even if robots can be engineered to fold laundry, the question of whether they threaten jobs, as a result, is more complex. Computer engineer Eric Holloway comments,
I’d wager most blue collar jobs are free from AI/robot competition. Only highly trained white collar workers need fear AI taking their jobs. In general, the more narrow and precise the training to do a job, the more likely an AI will take it.
That sounds counterintuitive until we think about it a bit. It’s one thing to engineer a robotic system for a narrowly defined job: Fold 1000 towels of a specific size, shape, and fabric. But a real-life load from a household dryer may contain 1) identifiable items that can be folded (towels, sheets), 2) identifiable items that can’t be folded (oven mitts, moccasins), and 3) unidentifiable items (coins, paper clips, brooches, loosened buttons, shreds of facial tissue). Engineering a system that can deal swiftly and correctly with a grab bag of “laundry” involves many more constraints.
As Holloway implies, folding laundry is a lower-status job. But it isn’t a low-skills job. The status of a job depends on many factors including, for example, the rarity of the skill. For example, the film producer can fold her own laundry if need be but the laundry service she uses while on location cannot replace her on the set. On the other hand, the monthly exchange of invoice and payment between the laundry firm and the film company could easily automate a white collar worker out of a job. Routine makes that skill easy to automate. Time will tell.
There is a model, Foldimate, that worked better at this year’s trade show.
But the firm admits it is struggling to keep the proposed cost and space requirements within average household budgets.
That, rather than AI limitations, may be the dealbreaker for consumers. Folding one’s own laundry at home does not require dedicated space. It uses only uncompensated time, perhaps while watching a movie. The profitable future for a “foldbot” may be commercial, industrial, and institutional use. In that setting, the dedicated space is factored into new construction or leasing and employee time really is money. How that affects jobs will depend on what, besides folding laundry, the employee is needed for.
At any rate, a handy device that folds our personal laundry isn’t just around the corner, it seems.
See also: Consumers were not buying robots as friends this year.