Your recycling is a mess. AI could help.
Recycling is hard. Most Americans don’t know what should go into their recycling bins - and when they guess, they often guess wrong and gum up the works of recycling facilities in the process.
This year, the United States will spend tens of millions of dollars to teach Americans the right way to recycle. But some companies and researchers think it would be smarter to outsource the job to computers.
Waste-sorting plants have been using machines to separate recyclables for years. But the machines - which use magnets, gravity, puffs of air and other rudimentary methods to sort through trash - aren’t perfect. And human workers can only sift through a fraction of all the recycling that comes through these facilities.
As a result, the “sorted” recyclables, particularly plastic, wind up contaminated with other forms of trash, according to Lokendra Pal, a professor of sustainable materials engineering at North Carolina State University.
“Today’s technology cannot really detect all the contamination that could be coming with those plastics,” he said. “But if you know what contaminants are coming, there’s a better chance you will be able to process it and get a cleaner product.”
That’s why Bollegraaf, the world’s biggest builder of recycling plants, and the AI start-up Greyparrot are rolling out artificial intelligence systems for sorting recycling. The companies plan to retrofit thousands of recycling facilities around the world with computers that can analyze and identify every item that passes through a waste plant, they said Wednesday.
“The destiny of a product is decided by this plant,” said Ambarish Mitra, who co-founded Greyparrot. “This is where the highest impact can be made - not in the bins, not in the household, not with consumer education - because this is where the decision is really made about what’s going to happen to that product.”
- - -
Putting a dent in recycling
Not much trash - and almost no plastic - actually gets recycled. About a third of U.S. garbage gets recycled or composted, according to the Environmental Protection Agency’s most recent estimate, which relies on 2018 data. For plastic, the figure is closer to 5 percent, according to a 2021 report from Greenpeace. The rest goes to landfills and incinerators.
Recycling rates are higher in Europe, where people are more willing to sort their own recycling into different bins for different categories. Americans have shown less enthusiasm for that, according to Barbara Reck, a senior research scientist at the Yale School of the Environment.
To make recycling easier, many U.S. cities don’t ask Americans to separate paper, glass, metal and the many forms of plastic. They just ask people to put anything recyclable into one bin - and let waste plants do the sorting for them.
But waste plants don’t catch everything. Greyparrot has already installed over 100 of its AI trash spotters in about 50 sorting facilities around the world, and Mitra said as much as 30 percent of potentially recyclable material winds up getting lumped in with the trash that’s headed for the landfill.
Failing to recycle means companies have to make more things from scratch, including a lot of plastic from fossil fuels. Also, more waste ends up in landfills and incinerators, which belch greenhouse gases into the atmosphere and pollute their surroundings.
Mitra said putting Greyparrot’s AI tools in thousands of waste plants around the world can raise the percentage of glass, plastic, metal and paper that makes it to recycling facilities. “If we can move the needle by even 5 to 10 percent, that would be a phenomenal outcome on a planetary basis for greenhouse gas emissions and environmental impact,” he said.
Cutting contamination would make recycled materials more valuable and raise the chances that companies would use them to make new products, according to Breck. “If the AI and the robots potentially helped to increase the quality of the recycling stream, that’s huge,” she said.
But she said the best way to raise U.S. recycling rates is to persuade more Americans to recycle. “What’s most important is volume,” said Breck, who helps lead the REMADE Institute, a research group backed by the Energy Department. “If somebody is not recycling at all and you can convince them to start, you get higher recycling rates.”
- - -
How does AI sorting work?
Greyparrot’s device is, basically, a set of visual and infrared cameras hooked up to a computer, which monitors trash as it passes by on a conveyor belt and labels it under 70 categories, from loose bottle caps (not recyclable!) to books (sometimes recyclable!) to aluminum cans (recyclable!).
Waste plants could connect these AI systems to sorting robots to help them separate trash from recyclables more accurately. They could also use the AI as a quality control system to measure how well they’re sorting trash from recyclables. That could help plant managers tinker with their assembly lines to recover more recyclables, or verify that a bundle of recyclables is free of contaminants, which would allow them to sell for a higher price.
Over the next few years, Bollegraaf and Greyparrot plan to retrofit thousands of material-recovery facilities with AI trash-spotting tools. Bollegraaf has built thousands of these facilities, including 340 in North America, which it says account for a majority of the recovery plants in the world.
Mitra said the trash-spotting computers could one day help regulators crack down on companies that produce tsunamis of non-recyclable packaging. The AI systems are so accurate, he said, that they can identify the brands on individual items. “There could be insights that make them more accountable for … the commitments they made to the public or to shareholders,” he said.
Related Content
Mattis secretly advised Arab monarch on Yemen war, records show
Back to the wall, Alejandro Mayorkas tries to build a border legacy
