Since mid-2016, Apple’s iPhone Photos app has used metadata analysis and image-recognition technology to sort users’ photos automatically and tag them for easy extraction. The categories include everything from various animals to inanimate objects like furniture—but as Twitter user @ellieeewbu noted on Monday, it also includes the archaic term “brassiere.”
In a limited straw poll, female Gizmodo employees universally confirmed suspicions this term is archaic.
Some users noted that Apple’s image-recognition technology seemed to be struggling to determine what should be included under the tag. According to one Gizmodo editor, searching for “brassiere” on her device results in “just pics of me in tanktops. I guess it thinks the thin straps are brassiere straps.” Others reported inaccurate results including a meme of Mona Lisa.
While Twitter users indicated searching the brassiere category did return photos of themselves in underwear, they also reported it identifying pics of themselves in swimsuits, and in a few cases nude or having sex (NSFW).
“Brassiere” is not the only similar term included on the list, but the Photos app appears to have specific, um, taste. Developer Kenny Yin posted a list of Photos’ 4,432 recognized scenes and objects in 2016 and sure enough, “brassiere” appears on the list—as well as “bra” and “bandeau.” But no other types of undergarments appeared to be featured, like panties, boxers or briefs, or even the term “underwear.” (Nor did terms referring to a number of unclothed body parts.) So I guess the iPhone knows what it likes.
It’s possible Apple has updated its list of recognized items in the meantime. As the Verge noted, the company says all object detection is done locally on the device and not transmitted to Apple servers, though users should still be wary about what they might be automatically uploading to their iCloud. Many of Apple’s competitors including Google and Facebook have similar object recognition tools.
We’ve reached out to Apple for comment, and will update this post if we hear back.