For original article, click here.
EVERESTLABS
AI “can” be a good thing. As in a robot that uses the technology to sort aluminum cans with at least 80% accuracy, picking out recyclables on a conveyor belt at a rate of 60 cans per minute.
“We tell customers we can guarantee 80 (percent), but we’re seeing north of 90% accuracy,” says JD Ambati, CEO and founder of EverestLabs in Fremont, California.
The five-year-old AI and robotics company is working with recycling facilities around the world to capture three times the recyclables as human sorters, Ambati says.
“A human being may sort 10 or 20 items a minute in an eight-hour shift. We have seen our robot grab as high as 40, 60, 65 picks a minutes in an eight-hour shift. It all depends on how much material is on the belt.”
One example is at a Caglia Environmental recycling plant in Fresno, California, which deployed EverestLabs’ AI and robot in September 2023. The facility is reportedly capturing 1,400 cans per day and on track to capture 32,000 pounds of cans per year, or more than 1 million cans annually.
EverestLabs developed RecycleOS for recycling plants, aka materials recovery facilities or MRFs. The software enables a robotic arm and 3D camera to recognize recyclables like cans and other items as they roll by on a conveyor belt, then quickly pick out the right items from the line.
This all means fewer items are going to landfills and more items are being recycled, reducing the greenhouse gas emissions needed to make new materials from scratch, Ambati says. The robot at Caglia was funded by the Can Manufacturers Institute, a trade association that’s sharing in the revenue from captured cans.
Caglia is just one example, Ambati says. Dozens of recycling plants/MRFs in the United States are using the EverestLabs system, including in New York and Michigan.
Other companies using AI and robots to improve recycling include AMP Robotics in Colorado and Pellenc ST in France. They’re all working toward a 100% by 2050 recycling target for the world’s aluminum cans set by industry leaders at COP28.
“That is a very good target and it’s a target that can be achieved,” Ambati says. “It can be achieved today because we have sophisticated AI robots and other automation solutions that are available in the market that can identify and sort accurately with the greatest degree of accuracy.”
Supporters of EverestLabs include Translink Capital and NEC Orchestrating Future, which led a $16.1 million Series A funding round in 2022, and investors BGV, Sierra Ventures, Morado Ventures and Xplorer Capital.
EverestLabs says its system was trained on about 4 billion images of recyclables.
EVERESTLABS
Two of its top executives come from the world of self-driving cars, Ambati notes. Just like a car that has milliseconds to react, the robots at Caglia and other facilities have milliseconds to identify and pick out recyclable cans and other items from the stream of stuff that people put into green bins.
“We are scaling significantly in the U.S.,” the CEO says. “We have dozens of customers today and we’re working on strategic international opportunities as well.”
He says accuracy rates of 80 percent-plus will grow to 98-99% by 2030 at the rate his company is going.
“You’re hearing all this noise about AI destroying the world,” Ambati says.
“This AI-that AI will write my thesis or create pictures and all that. You know what? What we’re showing to the world is a beautiful use case of AI actually advancing the planet, advancing the society.”
Humans will still be needed in the recycling world, he adds, but AI complements and increases the human quotient “to do more things, better.”
