As the internet gets more and more image-driven, it can be hard for blind people to participate in social media. But soon that could change, reports Cade Metz for WIRED. Facebook’s accessibility team is working on an artificial intelligence tool that can describe photos to blind people.
The tool uses the same deep learning techniques that Facebook already uses to identify individuals in photographs, writes Metz. When Facebook’s servers look at millions of photographs, they can start to recognize individuals. The same principle is used to help the AI describe photos, then feed those descriptions into a computer’s text-to-speech tool, Metz explains.
Earlier this year, Metz profiled Facebook’s accessibility team, which showed the huge demand for the image tool. At least 50,000 blind people currently use Facebook along with Apple Voiceover technology and many others use alternative text-to-speech programs.
The internet's heavy demand on vision leads to a myriad of concerns about its accessibility for people with disabilities. For example, as Instagram becomes increasingly popular, it’s leaving vision-impaired users in a lurch. Though blind people are able to take photos and use the social network, to actually 'view' photos they are at the mercy of the captions written by other users.
There are already some tools that can help take the internet from a visual to an auditory interaction. Text-to-speech programs can interpret emoji—which are described with terms like “grinning face with squinting eyes,” according to the BBC’s Ouch: Disability Talk podcast. But visually impaired people's experience of emoji is only as good as the captions themselves.
Bottom line: For the vision impaired, a transition to a more auditory internet can’t come soon enough.