In a working world of increasing speed and information overload, how do we keep our focus on the things that truly matter? Meet the machine learning team using cutting-edge technology to help make work more meaningful.
Technology may have made us more connected, faster and efficient. But it's also dramatically increased the volume of information we're expected to deal with. It's estimated that 120 billion business emails will be sent every day this year, while the traditional nine-to-five has become more like five-to-nine, as mobile makes us instantly available to colleagues around the world.
Imagine a universal personal assistant delivering what we need when we need it.
In the future, rather than increasing the amount of information we're expected to handle, technology will help to create more personalised experiences. It will strip away the noise leaving only what's important. Relevance will be king.
Dialling down the noise
This has long been the promise of AI. Forget the rise of the robots taking jobs from humans. Imagine instead a universal personal assistant, sifting through the detritus of emails, meeting invites, instant messages, notifications and project updates to deliver what we need when we need it.
Sounds far-fetched? If you're using Workplace, you've already experienced it.
Messages with higher relevance
The mission of the Workplace machine learning team is to make the platform as relevant as possible to every single person that uses it. "It means that when you log in to Workplace, it will help you find what you need to do your job properly. It means that it will be very easy for you to collaborate with others. And it means that your work will be better," says Tamar Bar Lev, the Engineering Manager who leads the team.
"It means that when you log in to Workplace, it will help you find what you need to do your job properly."
Machine learning is a subfield of AI that enables computers to learn how to solve various tasks by providing them with information, without the need to explicitly program them.
Bar Lev's team works on ranking and recommendation problems. Their job is to make sure that every time you launch Workplace, the algorithm brings you the posts and recommendations that you're most likely to find useful while "downranking" the ones you don't need.
Predicted probabilities improve the user experience
Advanced machine learning depends on raw computational power. "We train our models on previous examples," explains Bar Lev. "We feed our models with information on how users behaved in the past and how they interacted with different products. With enough information, the models are able to learn how to predict the probabilities of certain events. We can then use these predicted probabilities to decide which stories and recommendations are the best to show for each and every user."
"We combine all these probabilities into a final score. The post with the highest score is presented first."
This quiet rocket science kicks into gear every time you open Workplace. "For News Feed [the scrolling heart of Workplace that displays posts and recommendations], we go and fetch a load of posts from people that you follow, people in your team, from groups you're in," Bar Lev explains.
"Then we extract certain features for each post – such as how many times you clicked on a post from that user in the past seven days – then we feed it to our algorithm. That gives us certain probabilities: what is the probability that you will comment?"
"What is the probability that you will "like" it? What is the probability that you will react to it? We combine all these probabilities into a final score, and we sort all the posts in descending order. In the end, the post with the highest score is presented first."
Making work more meaningful
It might not sound like the AI Hollywood promised (or warned you about). But this is the cutting edge of machine learning, powered by years of experience from Facebook's consumer research. "When we do News Feed ranking in Workplace, we're not just starting from scratch," says Bar Lev. "We've learned a lot from many years of developing consumer News Feed in Facebook. We know what works and what doesn't. The challenge is fitting that into the work environment."
When you log back in to work the following morning, we want to make sure that you don't miss something crucial for your work.
Such as? "We want to make sure you don't miss anything important. If you finish your work day and log back in the following morning, or after a long weekend, we want to make sure that you don't miss something crucial for your work. Understanding which posts should be shown in such cases is a very different problem from understanding which posts should be shown to you in consumer Facebook every time you open the app."
Work graph and better Workplace connectivity
The exciting part is that these are still the early days of AI, and the team (like the algorithm) is learning fast. Where Facebook built people's social graph, an interconnecting network of friends, family and interests, Bar Lev's engineers are developing the work graph, which in its own way could be just as revolutionary.
"One of the questions we're looking at is: How do we make all of your connections more meaningful? We want to look at how meaningful it will be for you to follow someone. Will you meet with them afterwards? Will you attend the same events?
"When we look at your top colleagues, we want to see if you're collaborating with them on some Quips or G-drive. Are you checking in with them? Are you commenting on their posts? We want tounderstand your network so that when we suggest things or show you ranked stories, we know they'll be the ones that matter."
And then? "It means that your work will be better. You'll feel that without Workplace, you'll be slowed down. You will kind of… miss it."