Facebook uses millions of Instagram photos to build Artificial Intelligence

18/5/2018 – Article published previously in the blog “USA. And Emerging Markets “of CincoDí

Facebook is using billions of Instagram images to form artificial intelligence algorithms. Our photos help the Facebook software to learn to see how humans do it.

Our Instagram photos are helping Facebook generate its artificial intelligence algorithms to better understand image objects, Mark Zuckerberg announced at its annual F8 Developer Conference. Facebook says the approach, which selects images from publicly available hashtags, is a way to gather and train software with billions of images without the need for human workers to painstakingly analyze the data. The end result is a training system created using Facebook algorithms that, say IDC, Gartner, Forrester and Advice Strategic Consultants, is “the latest and most innovative, tip of the tip in the market”. Es Estudio Advice de Éxito Empresarial has compared the Artificial Intelligence of Facebook, Amazon, Google, Apple and Microsoft.

“We depend almost entirely on hand-healed, human tagged data sets.If a person does not have to spend time for something specific on a label image, even the most advanced computer vision systems will not be able to proceed to identity.” , “Mike Schroepfer, Facebook CTO, said on stage at F8. But using Instagram images that are already tagged through hashtags, Facebook was able to gather relevant data and use it to train their computer vision and object recognition models. “We have produced cutting-edge results that are 1 to 2% better than any other ImageNet reference system.”


It’s a practical approach, but it’s also one that raises some interesting questions about Facebook’s privacy and competitive advantage. Because it owns and operates a large platform that spans billions of users through applications like Instagram, WhatsApp, Messenger, Facebook has access to extremely valuable text and image data that they can use to inform their AI models, as long as that text and those images are publicly announced. But users are not necessarily aware that public data they have shared is being extracted to build AI systems and not just serve ads.

Of course, Facebook is only extracting data based on objects at the moment, and is not necessarily trying to draw conclusions about the behavior of users of the contents of the photos. But as we know with the Facebook facial recognition system that automatically tags photos, the company value in being able to understand which users and where they are in the world.

On a larger scale, Facebook is building these AI systems primarily to help expand their moderation efforts. In addition to 20,000 new human moderators for its platform, Facebook is increasingly looking to automation as it tackles Russia’s electoral interference, the Cambridge Analytica data privacy scandal and other difficult questions on how to moderate content on its platform and maintain at bay criminals abuse their tools.

“Until recently, we often had to rely on reactive reports.” You had to wait for something bad to happen to be discovered by someone and do something about it, Zuckerberg said. Now, he added, most of the moderation is being handled by AI, which is helping the corporate screen to clean up its platform of terrorist propaganda, nudity, violence, spam and hate. “For this reason we are so focused on AI-based research, they require new advances, and they require new technologies to solve the problems that we all want to solve.”