AI Recycling

AI recycling

Recycling is a good thing, but it’s not completely successful. Not everyone recycles, for one thing. Back when consumers had to sort their refuse into glass, cardboard, paper, and aluminum, a minority of people made the effort. Now that we just need to sort into recyclable and not-recyclable, more of us will take the time. But that doesn’t mean that all the discards get into the right place at the right time. AI recycling has the potential  to make the process work better.

Sorting takes place at the recycling facility, and it’s automated. Different materials behave differently, so conveyor belts can sort a lot of the items successfully. But not all of them. Estimates of the amount of refuse that gets shunted into trash piles instead of recycling range from 10% to 30%. Those numbers may reflect regional differences in what can be recycled as well as different rates of human error. Those human errors could include both poor communication of what can be recycled in a community and careless consumers.

Estimates of how much of the stuff that gets identified as trash when it should be recycled are even more varied. For-profit recycling centers usually use humans to ensure that their machinery doesn’t miss anything — or at least that whatever the machines miss get a second chance to make it into the right piles.

Entirely automated systems can miss items and misclassify things.

Automation vs. AI

Sorting garbage is certainly the kind of job that ought to be automated. It counts as dull and dirty for sure, and it is also dangerous. The Nation reports that recycling averages an 8.5% accident rate,  compared with 3.5% for industries generally and 5.1% for all waste management workers.

But it’s also the kind of job that’s hard to automate. Refuse is not uniform and streams of household waste don’t arrive in the recycling plant neatly arranged so that each item flows past the sorter in uniform positions. It’s a messy mishmash that robots can find hard to deal with.

Enter artificial intelligence.

An AI system can learn to recognize items much the way a human being can. AMP Robotics has developed a system that can recognize materials no matter how they’ve been deformed, by learning about specific packaging as well as the characteristics of particular materials. “My company now has the world’s largest data set of recyclable material images for use in machine learning,” the company’s owner said in an article in IEEE Spectrum.

With AI, human beings can be almost entirely removed from the job. Full automation becomes possible.

It is this kind of change that shows the full potential of AI.