Go ahead, hit that BUY NOW button. Procure that sweater or TV or pillow that looks like a salmon fillet. Hit that button and fulfill the purpose of a hardworking warehouse robot.
Just know this: the more you rely on online shopping, the more online retailers rely on robots to deliver those products to you. Robots shuttle cabinets of goods around warehouses. Other robots scan barcodes to do inventory. And, increasingly, robotic arms do what once only humans could: Sort through a vast array of oddly-shaped objects to compile large orders, all to be shipped to you, dear consumer.
“To my mind, the big story in 2017 has been an inflection point in e-commerce,” says roboticist Ken Goldberg of UC Berkeley. “Companies like Amazon and others are now delivering products at an unprecedented rate, something like 500 packages per second. And that is only going to grow.”
And evolve. Working robots no longer just lift heavy objects or weld or do other large, brute-force tasks. The new breed of robot rolling through fulfillment centers like Amazon’s is more advanced, more nuanced—and more collaborative. And while automating parts of these processes makes order fulfillment cheaper for e-tailers (and, consequently, you), it’s also fueling a robotic renaissance that will have implications far beyond the warehouse.
When we think of factory robots, we think of the machines doing the exhausting bits—like rolling around fetching items—while the humans do what they do best: manipulation. This paradigm continues to exist. A human remains in charge of the crucial (and surprisingly complex) final step of actually filling boxes because nothing can beat the dexterity of the human hand. For now, at least. The machines are making rapid progress on that front.
That’s due in part to Amazon’s Picking Challenge, in which teams put their manipulative robots to work. This has helped bridge a divide between academia and industry. “Robotics for the longest time has been really just about research, and not about putting things in the real world because it was too hard,” says UC Berkeley roboticist Pieter Abbeel, whose new company Embodied Intelligence is on a quest to make industrial robots smarter. “And I think the Amazon Picking Challenge is kind of one of those things where people are saying, Wow, this is a real-world thing, a real need and we can do research on this.”
At a San Francisco startup called Kindred, for example, engineers are teaching robots to do that final step of fulfillment. Using a technique called imitation learning, engineers steer the robot to show how best to grasp a wide range of objects you’d find at a marketplace like Amazon. “Some are soft and squishy, some are hard, some are heavy, some are soft,” says George Babu, co-founder of Kindred. “And there’s no way you can program that.”
Then a second technique, known as reinforcement learning, kicks in. The robot takes what it’s learned and through trial and error further hones it, both for speed and accuracy. Theoretically this would not only supercharge the fulfillment process, but make it more flexible. For instance if you’re a clothing retailer and winter rolls around, you’ll need to teach the robot to handle bulkier items like coats. (Kindred is running a pilot program at the Gap.) Why write out a bunch of complicated new code when you show the robot how to adapt?
But even in a relatively structured environment like a fulfillment center, the machines face plenty of obstacles. Some of them literal, like the humans they’re working with.
Robots Are Friends, Not Bullies
The need for increased collaboration between human and robot is forcing companies to look closely at how they integrate autonomous machines into the workforce. For Amazon and its 100,000 working robots, that has meant doing something very human: listening. Workers were a vocal part of the onboarding process. “Our associates actually got as granular as giving feedback on the fabric of the shelf and the color of the pods,” says Amazon spokesperson Nina Lindsey. “And that design has actually made it more efficient for our associates to find items.”
Which to a cynic might sound like workers willingly hastening the demise of their jobs. But in the near term, that’s not what’s going on here. Amazon has ramped up its hiring of humans right alongside its hiring of robots. And there is very much a place for humans as there is a place for robots. “Technology is extremely good at performing tasks that people do, but jobs are more than tasks,” says David Schatsky, managing director at Deloitte and co-author of a new report on robots in the workplace. “So jobs will change, but I don’t see a wholesale elimination of lots and lots of job categories.”
Still, the automation of jobs is nothing new. Consider that at the end of the 1700s in America, 90 percent of workers toiled in agriculture. Fast forward to 2012, and that number is 1.5 percent. Warehouse work is fundamentally different, but it’s not hard to see a time in the near future where increasingly sophisticated robots stop being collaborative and start replacing humans on the line. Whether that means those humans shift into more creative work, or they end up supervising the machines, will depend on the job.
So here we have the convergence of several factors that have kicked off a robotic renaissance. For one, the sensors that allow the robots to navigate a chaotic environment have gotten way cheaper at the same time as they’ve gotten way more powerful. Two, AI has vastly improved. And three, there’s money to be made—lots of it. E-commerce just keeps growing and growing, perhaps hitting $600 billion a year by 2020 in the US alone.
Which is not to say the e-commercers get to have all the fun. Expect the technologies developed for order fulfillment to spill out into the real world. The companion robots that have begun invading our homes will navigate better and better, taking a cue from their warehouse comrades. The machines will get all the smarter and easier for regular folk to teach, perhaps thanks to companies like Embodied Intelligence and Kindred. And that elusive dream of robotics, getting the machines to recognize and grip and manipulate a wide range of objects, could well come about because it’ll make someone in e-commerce a lot of money.
So go ahead, hit that BUY NOW button. The machines (and capitalists) thank you.