Sorting it Out

by Dan Morrell

Illustration by Michelle Thompson

In 2013, Matanya Horowitz, PhD (PhD ’14), realized that he had a solution in search of a problem.

“This groundswell in consumer interest is having a huge impact in the recycling industry, and when coupled with technology, gives me confidence that the world is going in the right direction.”

Horowitz was in the midst of his postdoc work at Caltech at that point, studying robotic control theory and path planning. In his spare time, he was researching artificial intelligence tools and building a deep learning “vision system” whose ability to identify specific objects actually improved with every new object it observed. Taken together, his pursuits could create a powerful robotic sorting system.

But where would such a system have the most impact? This kind of technology, he figured, could have applications in manufacturing, e-commerce, and even sports.

Eventually, though, he landed on recycling. It wasn’t a new idea— robotic sorting had been predicted as a solution since recycling was first popularized by the environmental movement of the 1970s. But even 50 years later, the recycling industry was still employing humans to separate out its aluminums, papers, and plastics, and staffing challenges were holding the industry back.

It made for a perfect fit: His vision system could detect recyclable materials by accurately identifying bits of brand logos and elements of shape—no matter how crushed the cans or cartons—and robotic picking arms could sort the materials. This process would not only improve the accuracy of the sorting process, creating more pure, higher-value materials, but it could also help solve the labor crunch. Recycling needed to be a better business, and his system, he figured, could make that a reality.

Eight years later, the initial promise that Horowitz saw is bearing fruit. His company, AMP Robotics, has 200 employees, 200 robots deployed, and $78 million in venture funding. AMP has been a finalist for Fast Company’s 2021 World Changing Ideas list and made the Forbes AI 50 list for the past two years; Horowitz was also named a 2021 Ernst & Young LLP Entrepreneur of the Year.

He not only found the right problem, but he found it at the right time. “As modern society has become more aware of its negative impacts, consumers are pushing for more sustainable approaches to business,” says Horowitz. “This groundswell in consumer interest is having a huge impact in the recycling industry, and when coupled with technology, gives me confidence that the world is going in the right direction.”

Rise of the Machines

 

Matanya grew up fascinated by robots, enchanted early on by Voltron and Transformers cartoons. But the reality of a career in the field didn’t cross his mind until high school, when he came across the DARPA Grand Challenge, a global competition that tasked teams with building and racing autonomous vehicles. Ultimately, none of the cars even finished the race, and Horowitz remembers being unimpressed with their construction. “They were clearly clunky,” he says. “It looked like they were held together by duct tape or something—and lots of wires everywhere.” But it wasn’t the tech that moved him, it was the young, scrawny grad-student types who built it. He had this idea that advanced technology was the dominion of secret, government labs. The reality, he saw, was that the future was being designed by people not much older than him. “And, I don’t know, are they any smarter than me?” he thought. “I should be able to do this.”

After earning four undergrad degrees and a master’s in just four years at the University of Colorado Boulder, Horowitz headed to Caltech for graduate school, impressed by the research and reputation of the faculty. Among them was famed robotics professor Joel Burdick, PhD, with whom Horowitz would ultimately work during his PhD studies. And while Horowitz’s work in robotics and control theory would play a major role in the sorting mechanisms that AMP’s robots employ today, it was a chance encounter that would round out the foundation of the company. As a hungry grad student attracted by the promise of a free meal, Horowitz decided to attend a campus lecture on deep learning, a type of machine learning that solves problems by breaking them down into a hierarchy of easier problems. He was floored. “I became convinced that this was going to be a revolution in computer vision,” he says. “And I wanted to be part of that.”

AMP officially launched in 2014, but the first few years were rough. After a long stretch of tinkering and development, AMP put its first robot in place in the beginning of 2016 at a facility focused on sorting out food and beverage cartons from the rest of the recycling stream. “It was a nightmare,” says Horowitz. “The thing didn’t work at all, it just immediately broke. It was very embarrassing.” He spent the next year waking up every day to new problems to solve—gummed-up suction systems, conveyor belt cleats ripping up his robots. “There was certainly a lot of self-doubt,” says Horowitz. “I think the question was, ‘Oh gosh, am I the guy who can figure this out?’”

Things started to take off in 2017. AMP sold five robots that year, started providing its AI to a Canadian recycling company, and raised its first venture capital money. Not only had Horowitz worked out some of the bugs in the system by then, but he had honed the economic pitch.

Recycling, he explains, is a tough business. In many cases, he notes, the costs of running a facility can exceed the price of the recycled commodities it produces. “But if you can bring the cost of sorting down, that can really make an impact. And that’s what robots do.”

Helping recycling plants sort more accurately not only creates higher-quality materials, he notes, but it can also help separate out more types of recycled goods—creating even more commodities to sell. “And if you do both, recycling starts to become this phenomenal business.”

Scaling Change

 

If recycling becomes a better business, Horowitz says, it will create more incentive to collect recycling. That, in turn, will create more recycling operations in areas of the country or even the world that haven’t had those options before. “In the developing world, where they don’t have waste management infrastructure, a lot of material is basically thrown in rivers and dumped,” says Horowitz.

But if these advanced systems like AMP’s start making it both simpler and cheaper to sort recycling, then “I think you can start to make inroads in this global problem.”

He hopes to watch this scenario play out over the next decade, with recycling becoming a growth industry and new entrepreneurs sprouting up around the globe, finding more ways to recycle a broader range of materials.

“I’m very much a techno-optimist,” says Horowitz. “What I see is that technology not only has the ability to make ‘doing the right thing’ easier, but can also make it the lucrative thing to do.”