Machine learning is at the heart of almost every product at Facebook. It powers the ability to help people discover new groups, products or brands they enjoy; connect with people across the world who may not speak the same language; and even power entirely new experiences such as 360 photos and 3D content. More importantly, it is used to help keep Facebook safe for over 1.5 billion people around the world.
Machine learning spans across all of Facebook’s apps, and engineers and researchers work cross-functionally with product teams to build exciting new features and tackle challenges that have never been solved before. At our global scale, it's an essential tool that can understand large volumes of information and connect people with what matters most to them, power new kinds of experiences, and help keep our communities safe. We spoke with engineering manager Daniel O. and research scientist Annie L. to learn more about the impact of machine learning at Facebook. Here’s what they have to say.
Building Communities
“Today, machine learning is being used in powerful ways to help us reconnect to our past on Facebook. It is the technology used to drive suggestions to pages you might like or people you might know. The News Feed team works on machine learning algorithms to surface the most relevant stories from people and pages in your connections. We do this by developing models that takes in personalized signals from your social graph. I also love that Facebook recommends groups based on common interests. Through Groups, Facebook is helping to build community across borders and providing a safe space for people to connect, share, and organize information.” —Annie L., Research Scientist, Machine Learning
“People who use Facebook come from a variety of backgrounds. They often speak different languages, come from all types of socio-economic backgrounds, and some have disabilities. What continues to amaze me is how Facebook uses machine learning to help people communicate with people they wouldn’t be able to connect with if this technology didn’t exist. We use machine learning to transcribe videos and create captions for people with hearing disabilities. We’re able to automatically translate text into the language of the person reading a post, so two people who don’t speak the same language can connect. Using object recognition technology, Facebook has created an “Automatic Alt Text” feature which uses artificial intelligence to convert text to speech and provide descriptions of photos for people who are visually impaired.” —Daniel O., Engineering Manager, Machine Learning
Maintaining Integrity
“Machine learning plays a pivotal role across Facebook because it enables multimodal content understanding. I lead a team of engineers that uses machine learning to proactively detect and remove policy-violating content across Facebook’s family of apps. Often times, when you visit other websites, you’ll notice ads of extremely low quality. This includes sensationalized headlines, or adverts for things like weight loss pills. These websites are helping advertisers monetize at the cost of their users. With machine learning, we’re able to address this on Facebook to make sure we’re giving people high-quality content relevant to them. ” —Daniel O.
“I work on News Feed, the first thing you see when you log into the app, and as we all know, first impressions can make a big impact. The Feed format was pioneered by Facebook and was a challenging engineering problem that we overcame. My team focuses on ensuring people see content that is relevant to them. We do this by developing models that control the order of the stories that appear, so people see the most interesting content first. Another big part of our work is integrity. We build models to identify and reduce the frequency of fake news, clickbait, and harmful content that can be polarizing to people. My team’s goal is to provide a positive and safe experience for people around the world who come to Facebook to discover content tailored to their interests.” —Annie L.
Fostering Innovation
“We use machine learning every day without realizing it, whether you’re connecting with your family using Portal, driving to the grocery store, or browsing Instagram. Machine learning is going to play a pivotal role in the future of how we communicate with others, and Facebook is leading the work in this field. We continue to grow and bring on engineers who are passionate about building for this future. As a manager, it’s my responsibility to ensure I’m helping engineers grow their skills, while empowering them to drive every part of a production cycle, from start to finish. Regardless of their level, engineers at Facebook have the opportunity to solve complex problems and truly own the work they are doing, from ideation to shipping.” —Daniel O.
“I'm switching apartments soon and currently going through the painful process of downsizing. I thought I'd give Marketplace a try and was excited by how smooth and efficient the process is compared to other similar sites! The machine learning algorithm correctly matches my listing to potential buyers and recommends similar items to the ones I've been browsing. The integration with Messenger and auto reply suggestions work seamlessly. The social aspects bring out more trust to the exchanges. And this is just one of the many products Facebook has been innovating to bring joy to people through AI.
As we move closer to Artificial Intelligence, it's an exciting time to be a part of this shift. Working at a company that is heavily invested in AI infrastructure and constantly exploring new ways to communicate and foster relationships through machine learning is inspiring. The opportunity to innovate and drive impact is limitless.” —Annie L.