South Florida Sun-Sentinel (Sunday)
AI spies our wasteful habits
In the United States, a third of food that’s grown is never eaten — and that’s too much
A hotel chain installs a camera in its trash bins to spy on what guests are tossing. Turns out its breakfast croissants are too big. Many are going to waste — along with profits.
A supermarket can suddenly see, hidden in its own sales data, that yellow onions aren’t selling as fast as red onions and are more likely to be trashed.
The brains behind both of these efforts: Artificial intelligence.
It’s part of an emerging industry that’s trying to cash in on a senseless human problem: The huge amount of uneaten food that goes from supermarkets and restaurants to the Dumpster. Much of that, if it’s not composted, ends up in landfills where it decays, sending potent planet-warming greenhouse gases into the atmosphere.
Enter a new business opportunity. A company called Winnow has developed the AI tool that spies on restaurant garbage. Another company, Afresh, digests supermarket data to look for wasteful mismatches between what a store is stocking and what people are buying.
In the United States, a third of food that’s grown is never eaten. Globally, 1 billion metric tons of food went to waste in 2022, according to the U.N. Environment Program.
Food waste accounts for 8% to 10% of global greenhouse gas emissions, roughly equal to emissions from aviation and shipping combined.
Adding to the problem: confusing “best by” and “sell by” labels that result in perfectly edible foods going into the trash.
Signs of progress are emerging from a group of supermarket chains that voluntarily pledged to reduce food waste in their operations in the western United States and Canada. Between 2019 and 2022, the eight chains that are a part of the Pacific Coast Food Waste Commitment reported a 25% decline in their total volumes of unsold food.
Afresh’s technology grinds around six years of sales data on every product in the fresh-foods section of a grocery store it works with. Its AI tool can divine when people buy avocados and at what price. It can mash that up with data on how quickly avocados spoil and in turn advise how many avocados to stock.
If Easter egg-painting season traditionally brings more egg sales, it can calculate how many more cases of eggs the store should order — and how many more bell peppers because shoppers usually make omelets with the extra eggs at home.
While an experienced store manager would likely know this, the AI would offer more precise information about many more products, Afresh co-founder Matt Schwartz said. It could recommend, for instance, that the manager order 105 cases of eggs the week before Easter, rather than 110.
“Every one case matters,” he said.
Winnow installs cameras above garbage bins in restaurant kitchens. The images are fed into an algorithm that can tell the difference between a half-pan of lasagna (valuable) and a banana peel (not so much). A group of Hilton Hotels that rolled out the tool recently learned that many of its breakfast pastries were too big and baked beans were commonly left unfinished.
Refed, a research group, found in its 2022 estimates that 70% of wasted food at restaurants is food that’s left on the plate, signaling a need to reconsider portion sizes.