What’s Makes Artificial Intelligence “Tick” and How it Aims to Revolutionize Waste Recovery Operations of the Future

Artificial intelligence (AI) has considerably evolved in the decades since the term was coined; the technology has caught on in industries like healthcare, finance, and IT, among others. Now the waste management sector is starting to pay attention to its capabilities. In this Q&A, Ambarish Mitra, co-founder and CPO of Greyparrot, discusses how AI can help solve long-standing, complex problems in this specialized sector.

Greyparrot, with $17 million in funding from Closed Loop Partners, Amcor, and Regeneration.VC, among others, developed and sells an AI waste analytics platform that operates in 10 countries.

In this Q&A Mitra keys in on specific problems AI is beginning to address. He touches on what’s hindering the accelerated adoption of new technologies and how to begin to push past the roadblocks. And he discusses Greyparrot’s platform, which categorizes 67-plus waste objects by parameters from financial value to potential emissions, to an individual product’s stock-keep unit (SKU).

Waste360: What are some pressing current challenges in waste management?

Mitra: The modern challenge is the sheer amount of material that’s being thrown away.

We produce just over 2 billion tons of solid waste a year. The World Bank estimates that by 2050, it’ll be closer to 4 billion tons. That’s concerning, because recycling rates for materials like plastic remain incredibly low. The majority of the objects we throw away end up in landfills, incinerators, or the natural environment.

Behind those growing mountains of waste is an unsustainable approach of production and consumption. Global circularity has actually decreased over the last few years, with some brands and packaging producers using more virgin materials than ever. It’s especially concerning when it comes to plastic, which is made using fossil fuels and emits 3.4% of the world’s greenhouse gases.

Waste360: What’s hindering the widespread adoption of innovations, and what’s needed to advance new technologies?

Mitra: There’s no shortage of ingenuity and passion in the waste management sector, but technology has often been implemented slowly and methodically. That’s understandable — recovery facilities are incredibly expensive to manage and adjust, and downtime to accommodate learning curves isn’t always feasible with the unrelenting influx of waste.

There is a tendency to hail one type of technology as the solution for everything. But different technologies and solutions, including artificial intelligence (AI) are going to need to come together.

New innovations, such as sorting machines, are often bolted onto the current process to make incremental improvements. However, this does not address the underlying issue of inflexible mechanical systems.

The waste crisis is developing faster than facility technology. Sorting remains a "black box": plant managers do not have full visibility into the sorting process. It’s a challenge that results in millions of dollars worth of valuable material being sent to landfill unintentionally.

Added to that is the fact that there isn’t much global alignment on policy (yet). That lack of consensus means that businesses interested in adopting environmentally-impactful technology are left without a blueprint. However, waiting for policy is a risky gamble. The waste management industry can thrive by striving for best practices and actively influencing policy. By taking a proactive approach, the industry can meet the challenges of modern waste management and ensure its long-term success.

Waste360: Tell us about Greyparrot’s platform.

Mitra: Greyparrot’s core product is an AI-powered waste analytics platform. We install AI waste monitoring units above conveyor belts in resource recovery facilities, which use cameras to capture images of waste objects.

Our AI model uses those images to categorize each object based on material, mass, financial value, potential emissions and brand — right down to individual stock-keeping units (SKU, e.g. Diet Coke, 300mL).

Our units send that data to an analytics dashboard in real-time, which facility managers use to optimize their sorting processes and gain a deeper understanding of material flows.

Greyparrot also provides a flexible vision system that integrates with existing and new sorting machinery through an open application programming interface (API). Our API provides the ability to embed cognitive intelligence into existing software and hardware frameworks. This technology is powering the flexible sorting plants of the future that can adapt to fluctuations in feedstock.

Waste360: What problems are you trying to solve for?  

Mitra: A lack of waste composition data means that operators aren’t getting an accurate look at their waste streams; this causes a number of problems our system helps overcome:

Waste360: What are some of the 67-plus waste categories your platform recognizes?

Mitra: Those would include categories of plastics; fibers; metals; and general object groups like batteries, waste from electrical and electronic equipment, face masks, fabric, and more.

Plastics are the biggest category of Greyparrot’s waste recognition library, largely because it has presented such a challenge. Greyparrot can accurately distinguish categories including standard polyethylene terephthalate (PET) bottles to food- and nonfood-grade material, black and colored plastics, different forms of high-density polyethylene (HDPE), and several common types of flexible film.

Greyparrot AI can also track a waste object’s brand, right down to the stock-keeping unit (e.g., Diet Coke, 330 mL).

Waste360: Can you discuss some key areas you believe recyclers must focus on for true impact that AI can help with?

Mitra: 1)Maximizing recovery and revenue. In order to have a functioning circular economy, we need a steady supply of high-quality secondary materials. The more we recover, the less consumer packaged goods companies (CPGs) will have to rely on virgin materials that threaten circularity.

Increased recovery also means fewer waste objects being landfilled, incinerated, or reaching the natural environment. Plastic, in particular, creates a “negative feedback loop” if it’s mismanaged — releasing methane and ethylene that warm the planet and accelerate the breakdown of further plastics, all while harming marine carbon sinks like phytoplankton.

2)Enabling quality control. The quality of secondary resources is directly tied to their price. Unless waste streams are completely homogenous, it has traditionally been difficult to track and guarantee the quality of recovered materials, which reduces their market value.

Regulation is tightening around secondary resources, with regions like the E.U., for example, now requiring food-grade materials to be at least 95% pure. Product quality will increasingly determine whether a bale is accepted or rejected, as well as define its price.

3)Impact on product design and regulation. Packaging producers have never had comprehensive insight into their products’ “end-of-life” phase. Along with production and transport, it’s the phase responsible for the most environmental damage.

Producers are rarely able to tell whether their products are recyclable in practice, even when using technically recyclable materials like TetraPak. By tracking brand and product data (leveraging AI), we’re able to show CPGs exactly what happens to their products when they reach sorting facilities. Brands can then design packaging with real post-consumer recycling data, making an impact at the very top of the value chain.

4) Efficient data collection. The waste sector contends with a global labor shortage. It’s prohibitively expensive to use manual sorters to collect the amount of data that’s possible with AI, but fluctuating purity means gathering that data is crucial.

We ran a comparison (based on U.K. wage data, and our own licensing model), and found that manually gathering data on 15 tons of PET would cost $5,500, compared to just $21 with automation.