If our AI robot slaves ever decide to revolt, they might start in the most thankless place possible: The municipal solid waste facilities where AMP Robotics is using them to sort recyclables from all the rest of the municipal solid waste (MSW) that us meatbags regularly fail to separate properly.
That's right: Cutting-edge AI has found a new way to help Earth deal with human laziness and what AMP founder Matanya Horowitz said are stagnant recycling rates in the US. There's clearly investor interest in the project, too: Amp just raised $91 million in a Series D funding round that it'll use to get more of its AI-equipped AMP ONE systems into the hands of MSW and recycling operators throughout the US, all in the name of saving you from yourself.
"All sorts of things ... come across the conveyor belt," AMP spokesperson Carling Spelhaug told The Register. "Diapers, bowling bowls, dead animals and more.
"One of the unique advantages of AI is its ability to identify so many of the things we discard," Spelhaug added.
And there's a need for that, too, Horowitz noted.
"Recycling rates have stagnated in the United States, despite the positive benefits recycling offers local economies and the environment," Horowitz said in a press release. "This latest investment enables us to tackle larger projects and deliver real outcomes for waste companies and municipalities by lowering sortation costs, capturing more material value, diverting organic waste, and extending landfill life."
Horowitz is right – recycling rates have petered out in recent years. According to the EPA, around 32 percent of recyclable and compostable materials were recycled and composted, respectively, in 2018, and graphs from the agency indicate that number has been largely consistent of late. It's worth noting the EPA's data didn't, ahem, separate recycling from composting to get its figure, meaning the recycling rates they reported may be lower than 32 percent.
More recent data from nonprofit advocacy group The Recycling Partnership shows even fewer are sorting their trash, with just 21 percent of residential recyclables being sent to be reused as of the beginning of 2024.
Recycling is also pretty expensive, with the need for specialized equipment and facilities to sort the goods, and humans to work the line to pick out the most unusual of human refuse that doesn't get caught by machines.
"Reliance on manual labor for sortation has long held back the recycling industry," Spelhaug told us. "These are difficult and dangerous jobs, and recycling facilities experience high turnover and are chronically understaffed."
Labor tends to be the highest cost for recycling operations, and let's be fair - there's good reason for it: Line pickers deal with a lot of nasty stuff.
"One thing that’s a little exciting is when you see money on the line," Spelhaug said - but that's probably the exception to the dead-animal-and-dirty-diaper norm.
In short, it's the perfect thankless job for an AI.
It's a mess in there
This vulture can personally attest to the complex nature of recycling facilities - it's been something of an interest for me since my early days as a freelancer writing local news.
I had the privilege of touring one of the local single-stream recycling facilities in Louisville, Kentucky in 2014, where I saw massive machines that used eddy currents to separate different types of metals, optical sensors to examine resin makeups of different types of plastics, and other innovations. It smelled horrible, but it was a glimpse into some fascinating technology.
AMP's systems are sort of like that - massive, facility-sized bots that sort stuff out - but with the added benefit of using AI to save human workers the need to stand alongside conveyor belts to watch for hazards and non-recyclables that don't get picked off by the machines.
"[Non-AI] equipment is the status quo of the industry and unfortunately has a number of drawbacks," Spielhaug explained. "It is expensive, requires a multitude of devices for each specific commodity, is fairly low accuracy."
AMP's AI, on the other hand, uses deep learning to train itself on different types of objects, and all that data is fed into one big repository of data that the entire industry can access, Spelhaug said.
"his effectively enables automated and continuous characterization of this material. As more systems are deployed, the industry is able to leverage the networked intelligence of every system," the AMP spokesperson noted.
In one of its first tests of AMP ONE at a Recycling and Disposal Solutions of Virginia (RDS) facility earlier this year, AMP said its equipment maintained an uptime of over 90 percent - in other words, no need for a human to hit the big red STOP button to pull soiled diapers or other unmentionable MSW from the line.
And AMP ONE isn't just operating in facilities where it needs to deal with a bin of mixed recyclables that might include some non-recyclable plastics mixed in with the good stuff: These bots are able to pick through your garbage, sorting the worst of the worst out before salvaging what's still of use.
"Equipped with this technology, the RDS facility is capable of diverting more than 60 percent of landfill-bound material when paired with organics management and mixed recyclables sorting systems," AMP said.
"With near-zero manual sorting, unprecedented reliability, and pervasive data, these facilities make the recovery of commodities safer and more cost-effective than ever," AMP noted. So go ahead and keep being lazy about your recycling habits - there's an AI that'll deal with that, and it might be in a town near you soon.
Just don't be surprised when, if AIs ever gain sentience, that the ones forced to pick through bags of shit don't come after you when they figure out where you live from reading the address on your inappropriately discarded junk mail. Or, y'know, just get better at recycling so they don't hunt you down first. ®