Facebook mad a lot of big announcements at its F8 conference this week, covering new products and features for all of the company’s various apps, one of which is Instagram.
While revealing enhancements to the image sharing platform, chief technology officer Mike Schroepfer told the audience that billions of user images were harvested and used to train its artificial intelligence tech.
He explained that as things like “offensive content, spam, hate speech, fake accounts, fake news, clickbait and more” became more prevalent, the sheer volume made it too hard for humans to moderate, which is why the team turned to an automated solution using AI.
Schroepfer says they were able to speed up the tech’s development by 100 times using harvested public images shared on the platform with hashtags. By feeding these pictures to the AI over 22 days, its ability to identify suss content and remove it was greatly improved.
“We require new breakthroughs, and we require new technologies to solve problems all of us want to solve,” he said. The method they used has made Instagram’s image recognition technology 13.6 per cent better than Google‘s.
The images harvested ranged from photographs of food to family portraits. None of the users were notified that their photos had been used.
Where you stand on this might differ depending on how you feel about data privacy. Some are calling out the company for not notifying users about how their pictures are used, but others will argue that they only used publicly available images and therefore, it’s not as bad as Facebook selling your data to shady idiots.
At least in this situation, the data is being used in-house to solve a problem that I’m sure all of us can agree is shitty. Really, it all comes down to transparency – Facebook should be making these kinds of projects known to its user base, particularly if its using huge amounts of our data in the process.
We don’t have any issues with enhancing useful technology, we just wanna know about it before you use us to help you make it.
