Facebook’s AI can work out what’s in your photos
Everyone’s favourite social network learns to tell selfie from sunset
Fear that Facebook might know a little too much about what you got up to last Friday night? Its latest photo-proto will do little to allay that disquiet.
Taking a leaf out of Google’s snap-sensing Photos tech, Facebook’s CTO Mike Schroepfer has outlined the Cambridge company’s latest foray into user upload management at Web Summit.
And this ain’t no binary robo–scanner: a demo illustrated a Newsfeed photo of a skateboarder being recognised by the system, and described in everyday words what it saw – including objects and what it thought they might be doing, in this case “a skateboard, a man” who was “riding”.
With real-world events understood in this way, the possible uses for the social media giant are pretty clear. Want to filter your feed to show only cat photos? Look no further. Sick of seeing Spurs fans celebrating? Facebook’s got the tools.
Well, it might have. Zuckerberg’s baby said it’s planning to push the AI team’s research at a conference next month and is looking into its sensing smarts as a potential future tool – which, with it apparently being 30 percent faster than most industry competitors, doesn’t seem such a long shot.
Google’s AI integration revolutionised library management with the simplicity of search offered by Photos – with an end to tedious tagging and several minutes of scrolling as you search for that embarrassing snap of Dave last Christmas.
Could Facebook’s tech do the same? It seems the company is keen to improve its search functionality, given the recent expansion of its reach to cover all historic public posts – so a future of auto-detected, searchable word labels for your holiday shots could well be on the cards.
[Source: The Next Web and Facebook]
Chris Rowlands
Freelance contributor 
