Critics dismissed early attempts to create AI-powered electronics as absurd. However, this AI gadget-in-the-making is about trash: Finnish company Binit tracks residential waste using LLM image processing.
Entrepreneurs such as Greyparrot, TrashBot, and Glacier have been working on AI to classify trash in order to improve municipal and commercial recycling. However, Binit creator Borut Grgic believes domestic garbage tracking is unexplored.
“We’re producing the first household waste tracker,” he says in Eltrys, comparing the AI device to a sleep tracker for garbage. The camera vision technology includes support for neural networks. We’re using LLMs to recognise domestic waste.” 
The pandemic-founded firm has raised roughly $3 million from an angel investor to make AI gear that looks good and lives in the kitchen, affixed to a cabinet or wall near the trash. The battery-powered device utilises cameras and other sensors to detect approaching individuals and scan items prior to disposal.
Grgic adds that they use commercial LLMs like OpenAI’s GPT for picture recognition. Binit records everything the household throws out and provides statistics, feedback, and gamification via an app, such as a weekly trash score, to encourage customers to minimise their waste.
The researchers tried training their own AI model for rubbish detection but got 40% accuracy. So they adopted OpenAI’s image recognition. After incorporating the LLM, Grgic claims over 98% garbage identification accuracy.
Its creator has “no idea” why Binit works so well. OpenAI’s ability to recognise rubbish may be due to its large training data set or its ability to recognise many objects. “It’s incredible accuracy,” he says, noting that OpenAI’s model’s great performance in testing may be due to “common objects” being scanned.
“It can even tell, with relative accuracy, whether a coffee cup is lined because it recognises the brand,” he says. “So basically, the user passes the object in front of the camera. Thus, they must temporarily stabilise it in front of the camera. The camera captures the picture from all angles at that moment.”
Binit uploads users’ waste data to the cloud for analysis and feedback. It plans to provide free basic analytics and subscription-based premium services.
The business wants to become a trash data supplier, which may be useful for the packaging company if it can grow.
Do consumers really need a high-tech gadget to alert them that they’re throwing away too much plastic? Don’t we all know what we’re eating and need to reduce waste?
“It’s habits,” he says. I suppose we know, but we don’t act on it.
“We also know it’s good to sleep, but when I put a sleep tracker on, I sleep more, even though it didn’t teach me anything new.”
According to US studies, Binit claims that its garbage transparency reduced mixed bin waste by 40%. Transparency and gamification may help individuals change behaviours, it says.
Binit wants the app to provide analytics and insights that reduce waste. Grgic plans to employ LLMs to provide location-based recommendations for the latter.
Each scanned piece of packaging, such as a plastic bottle, prompts the programme to create a card indicating its disposal. These local options might help you cut down on plastic, he says.
He also considers cooperation with food waste reduction influencers.
Grgic calls the product “anti-unhinged consumption,” another surprise. The startup supports sustainability awareness and action. To protect the environment for future generations, we must replace our throwaway culture of single-use consumerism with thoughtful consumption, reuse, and recycling.
“I feel like we’re on the cusp of something,” he says. “I think people are starting to ask: Is throwing everything away necessary? Can we consider fixing and reusing?
Why not simply use Binit as a smartphone app? Grgic says it depends. He adds that some families prefer a hands-free garbage scanner, while others prefer using a smartphone when dinner prep is messy.
Additionally, their app will enable free scanning, so they will give both alternatives.
The business has piloted its AI garbage scanner in five US cities (NYC, Austin, Texas; San Francisco; Oakland; and Miami) and four European cities (Paris, Helsniki, Lisbon, and Ljubjlana, Slovenia, where Grgic comes from).
He expects an autumn US commercial debut. He calls the $199 price range for AI hardware the “sweet spot” for smart home devices.
