We’ve been hearing a lot about the dangers of artificial intelligence, or AI. But many people are using AI to address some of our most pressing challenges, like climate change and managing waste.
Ian Goodine and Ethan Walko are two young people doing just that. Together, they founded the Somerville-based startup rStream Recycling and developed AuditPRO, an AI program that uses thousands of images of trash to train the program to sort waste more efficiently than humans.
Goodine and Walko joined GBH’s All Things Considered guest host Judie Yuill to discuss their work and the future of using artificial intelligence to effectively sort waste. What follows is a lightly edited transcript.
Judie Yuill: Ian, let’s start with you. Can you give us a rundown of how AuditPRO works?
Ian Goodine: The AuditPRO is essentially a fancy photo booth for trash. A user would walk up to it like they would any [dish drop off area] in a high-traffic venue. They’d place their waste in the system on a conveyor belt in this case, instead of directly into the bin.
That conveyor belt would magically turn on because it has detected the presence of an object, and we would be able to start taking pictures and videos of the objects that were thrown away, which are then used for analytics. We can use our AI algorithm to determine what was thrown out.
Yuill: Now, it doesn’t exactly sort the trash itself, at least not yet. Is that correct?
Goodine: Exactly. So, at rStream, our long-term vision is to develop a robotic sorting system. But as a startup, our phase-one pilot at UMass Amherst is this audit process system to ensure that the brain of the system—the artificial intelligence—is extremely accurate, so when we sort, things [will be] going in the right place next year.
Yuill: Ethan, how did this idea get off the ground? I understand it was part of a semester-long project while you were students at UMass Amherst.
Ethan Walko: Yes, that is correct. rStream began at UMass Amherst while Ian and I were doing mechanical engineering degrees. We had an opportunity to investigate recycling and the challenges through a senior design project, as you said, and then eventually it morphed into its own animal. And that happened during our graduate studies as well.
Then, this project, the AuditPRO, was derived from the desire to validate some of the data that we had been seeing for ourselves. In some past work, we performed manual waste studies at college campuses, and we realized how difficult that process is, but also how valuable it is to get that granular information that’s verified. The AuditPRO was able to do that automated and continuously without the mess and the struggles that come with manually sorting through hundreds of pounds of trash.
Yuill: Ian, AI can be used in many different ways. What made you want to focus your efforts on recycling and waste?
Goodine: There’s a ton of excellent research in AI, and many of those are going in different directions. But recycling and waste are some of the hardest challenges that we face as a society. But also, it’s a very hard technical challenge from the perspective of a computer that’s trained to generate decisions about what it’s seeing.
It’s very difficult to accommodate the amount of variation in the waste stream. Every can may look slightly different. There are tons of different types of waste fractions. Whether they’re cans or bottles or fibers, you name it, it’s there. This makes it a very exciting technical challenge for AI and an exciting place for us to pursue.
Yuill: Ethan, this project is still in its early stages, but you got federal funding from the U.S. Department of Energy. Give us your reaction to that, and share with us what the future has in store and what your vision is for this project.
Walko: Yes, gladly. The Department of Energy funding is very exciting, not just because of the cash, which will allow us to perform research and develop the technologies, but because of the validation that it comes with. That review process was conducted by a panel of experts in waste recycling and technology, so their assessment of our project as viable and exciting is huge validation for rStream, our mission and our technology.
What’s to come is we’ll be using this funding to deploy and develop the sorting system, which will use the brain from the AuditPRO, and we’re hoping to deploy that in the phase-two pilot, which will commence in 2025.
Yuill: What would you say to people who have only considered the negative aspects of AI?
Walko: I think that there is a lot of potential for AI if it is used responsibly. I think it’s an important conversation to talk about how we control the biases that can be present when artificial intelligence algorithms are learning. But I think that focusing on the negative side is limiting the potential and the many positive ways in which this technology can be used.
Yuill: Ian, what’s your take on that?
Goodine: The point of technology, in general, is to enhance the human experience. I can think of no better calling for technology than to eliminate miserable jobs like the task that people who could be doing more productive things with their time than sorting or facing anxiety about how they can dispose of things correctly.
This is the opportunity that artificial intelligence and corresponding hardware can enable, and we’re super excited to be at the forefront of developing that.
Yuill: Do you have a patent for this?
Goodine: Yes, we are U.S. patent holders for sorting and the AuditPRO system, and we have additional intellectual property around the proprietary nature behind our AI algorithms and the methods we use to create them.
Yuill: You’re both pretty young people, but you’re looking at something that potentially could be a real livelihood for you and employ others.
Goodine: Yes, we’re super excited that it’s been able to receive enough funding and attention that it’s funded not only us but a team of six technical people who otherwise may be working on other applications of their skills. To have them inside of a clean tech realm—and not only a clean tech, but an underfunded space inside of that, which is waste and recycling— is really exciting for us, and it creates a bridge into the future for this industry.
Yuill: I know you’re using it now for waste and recycling and to identify and sort trash. Can it also be applied further along the trash-recycling journey?
Goodine: Absolutely. This is partially what the Department of Energy, through the Riemann Institute, has funded. They’re interested in finding low-resource approaches to sorting. This could be a pipeline for future coverage centers in places that otherwise couldn’t justify the capital expenditure of a massive plant. So, figuring out how we could exist later on the recovery journey for these commodities—whether it be domestically or internationally—is a huge market opportunity that we could create for ourselves.
Yuill: We always like to ask everyone: Is there anything I didn’t ask you that’s pertinent that you’d like our listeners to know?
Goodine: One thing that is important for us is to point out that we’ve come so far with such a sparse set of resources. This being our phase-one pilot at UMass with the AuditPRO, we’re thrilled to have delivered that in a timely manner with a tight team and with the expansion of our resources through equity fundraising and through non-dilutive sources, such as the DOE and [National Science Foundation] grants we’ve received.
We’re really excited to continue pursuing our long-form mission of robotic sorting as quickly and effectively as possible so we can start creating benefits today.
