"Your food videos suck." — Meta's AI, probably
A Monday news bite.
Instagram’s parent really, really wants to make Reels — its short, TikTok-like videos — better. And it might have found a scalable way, as first reported by.
Meta filed a patent application for technology that can predict the quality of a video before it’s published, and it uses artificial intelligence, because of course it does. As described in the filing, the company uses machine learning trained on other videos to discern a video’s quality. “Quality” is vague here; it could mean videos poised to get the most likes or comments, videos shot with the best possible visuals in high definition, or great audio.
Its usage is twofold: first, video producers can instantly find out if their video is “good enough” based on any number of standards Meta decides are important. But also, Meta (or, more specifically, Instagram) gets immediate insurance against hosting awful content that might mistakenly go viral. Or, as the patent filing puts it, “…uninteresting content can undesirably proliferate on a content sharing platform. Such proliferation of uninteresting content can frustrate viewers and undesirably result in poor user experience.”
This presumably applies to all sorts of video content; but food videos are important enough to the company that they’re used as examples in the filing: