Kindred Robots Are Learning to Grab and Sort Clothing in a Warehouse for the Gap

Next time you shop online for jeans or T-shirts from the Gap, you may have artificial intelligence to thank when the correct items arrive on your doorstep. Kindred AI, one of our 50 Smartest Companies of 2017, has been testing Kindred Sort, its first production model robot, for the past six weeks at a warehouse as part of a new collaboration with the clothing company.

By pairing AI with remote human operators, the Kindred trial is using the operators’ skills to further train the machines, while also allowing  a small team of people to run operations around the world. For the Gap partnership, six human pilots located in Toronto are assisting robots based in Tennessee.

Kindred has made a pivot since we talked to cofounder and CEO Geordie Rose in March, removing the virtual-reality component from the human controller’s arsenal. Instead, the human members of the team are using a 3-D mouse paired with a standard keyboard to perform tasks that are currently too cumbersome for the robot to perform alone.

In this case, that task is grasping clothing, shoes, and accessories of varying shapes in the company’s warehouse. The robots are using deep learning and reinforcement learning to figure out how much pressure they should use when grasping and how best to grab the various items in inventory.

Kindred is charging Gap for the partnership, but not for the individual machines. Instead, the fee is for the robots’ time. As George Babu, Kindred’s chief product officer, told The Verge, “It’s pay-per-intelligent action. We’re not selling the robots, and it’s not a monthly lease either. We’re trying to create the AI, so we price the use of the AI, and the hardware price is baked into that.”