Image credit: Greyparrot
The plethora of plastic pollution plagues the planet. The damaging effects of discarded plastic items can be found in the environment and wildlife. Despite efforts to reduce and recycle waste, the Environmental Protection Agency notes that 75 percent of plastic products found in municipal solid waste isn’t even recycled and instead winds up in landfills.
Scientists at the University of Chicago’s Pritzker School of Molecular Engineering hope to combat that problem at the source. Researchers are using machine learning and artificial intelligence technologies to design plastics that decompose easily, with the end goal of reducing pollution.
Designed to Degrade
Products such as plastic bottles are made from polymers, which have amorphous and disordered structures. The shape, sequence, and structure of a polymer determines its characteristics, such as its strength, flexibility, and response to temperature. Due to the sheer number of possible structure combinations, random or trial-and-error testing would be inefficient.
That’s where machine learning comes in. One of the benefits of AI is that it processes information in a fraction of the time it would take a human. By running millions of computations quickly, researchers can identify the most promising potential structures and create models of them to test.
The UChicago team is using AI to establish the computational structures of nearly 2,000 hypothetical polymers using neural links that understand a polymer’s properties. The AI system uses that data and sequencing information to predict the polymer’s characteristics. The researchers found that “deep neural networks inexpensively and reliably predict structural properties with limited sequence information as input.”
The AI-based predictions enable scientists to create polymers and plastics with specific attributes, in this case, one that might hold water, but begin to decompose when left in the elements or in a landfill. In addition to developing degradable plastics, the UChicago team is exploring polymer combinations that would result in a strong, durable, and lightweight plastic that could be used to replace metals in aircraft and satellite fabrication.
AI Improves Recycling
Molecularly degradable plastics may one day reduce the amount of plastic in landfills, but it’s not yet ready for commercial use. A more immediate impact can be seen in waste sorting and recycling facilities, some of which use AI.
One example comes from Greyparrot, an AI technology company based in London. Greyparrot uses AI-based computer vision software to identify recyclable items amid the waste on a conveyor belt. It is faster and more accurate than the humans it works alongside, improving recycling rates. Currently less than 1 percent of waste is monitored, but automated waste analysis systems could bring that to 100 percent.
Greyparrot’s solution uses a monitoring unit, AI models, and a customizable dashboard that presents data in real-time. Using deep learning and AI technologies, the system analyzes and processes visible image data as the waste passes the monitoring unit. The software can recognize more than 40 types of waste materials and can pick out specific materials even in mixed waste streams.
Because it also can determine the composition of items, it can identify which items can be resold instead of going to the landfill. That means it can not only increase recycling rates, but it can flag items of value, such as scrap metal, which can be sold for a profit. As companies tap into the power of AI, they can begin to design waste out of the economy and help reduce our impact on the environment.