One company in the US state of Colorado has a new lesson for artificial intelligence. And it could help us end the drudgery of sorting our trash.
CGTN’s Hendrik Sybrandy has more.
It’s challenging and repetitive work performed in a noisy environment. These sorters help keep the lines at Alpine Waste & Recycling moving.
“The people we have are great at what they do,” said Brent Hildebrand, Vice President of Alpine Waste & Recycling. “They sort really well.”
Alpine’s Brent Hildebrand says that goes as well for one of his newer employees, six months on the job and getting better all the time. He is Clarke the robot.
“So Clarke, as he’s working right now, he’s watching everything that’s on the line,” said Matanya Horowitz, CEO of AMP Robotics. “And as he sees more and more material, he learns more and more.”
Clarke is Horowitz’s creation. The founder of AMP Robotics in Colorado used a National Science Foundation grant to build a device that uses artificial intelligence to pluck recyclable products from a conveyor belt. A recycling industry group helped sponsor Clarke’s effort to identify milk and juice cartons, something new technology now allows robots to do.
“You show them thousands of examples of bottles and cans and everything else and they begin to learn what distinguishes them, just like a person might,” said Horowitz. “So, it’s learning certain logos, certain shapes, certain textures are associated with one material over another.”
It’s not an entirely new idea. The movie robot WALL-E once confronted massive piles of garbage as the last robot on Earth. Other companies have also employed waste-sorting machines. AMP Robotics said its computer vision and machine learning cuts sorting costs. And robots come in handy when labor is scarce.
“It’s getting harder and harder to find people and when you do find them, you have to pay them more and more to keep them,” said Hildebrand.
Cartons are Clarke’s focus now but it won’t be long before robots are sorting plastic products and things like construction and electronic waste.
For now, the focus is on improving the software and reducing Clarke’s error rate. His grip still needs some work.
“We’re figured out the basics but there’s always going to be room for improvement,” said Horowitz. “We have a long road map of additional features and additional capabilities that we’re adding.”
Hildebrand said he sees improvements on a weekly basis.
With a growing world that keeps generating more of this stuff, job security may not be an issue for Clarke and all the smart sorters who are sure to follow.