Google’s Weirdest AI Dataset Yet: Its Own Garbage

Google has partnered with Mill, a company that produces a high-tech trash bin, to train AI algorithms capable of identifying food scraps. This initiative utilizes a dataset Google created from its own food waste, which was previously labeled and annotated for computer vision research. Under the agreement, Mill will also gain early access to unreleased versions of Google's flagship Gemini AI models and a team of AI engineers. Mill's smart bin processes food scraps into grounds for soil or animal feed, such as chicken feed. The technology aims to help businesses, like Whole Foods, make better procurement decisions by analyzing what food is being discarded. For instance, if a company consistently throws away macaroni and cheese, they can order less. This data could also help route surplus food to food banks and document donations for tax purposes. The project originated from Google's X moonshot division, known as Project Delta, which aimed to create an "air traffic controller for food" to match supply with demand. Although Project Delta was folded into the broader company in 2020, the food waste data was repurposed for this new partnership. Google's venture arm GV, GV, has supported Mill since its inception. The partnership is part of Google's AI Futures Fund, which connects startups with Google's AI tools in exchange for feedback. The article concludes with a quote from Mill's co-founder, Matt Rogers, expressing a vision where garbage collection happens less frequently, such as once a month, rather than daily.

(Source：Forbes)