Facebook is turning its mountains of data and computing power into artificial intelligence to make the platform as relevant and engaging as possible.
"It's about making sure that every moment you spend online, you spend it with the content you want to see, with the people you want to share," said CTO Mike Schroepfer at Facebook's annual F8 developer conference Wednesday.
The implications for brands buying ads on Facebook are massive, said developers.
"If there is something that is very different about Facebook in general, I would say it's the human aspect," said Lorenzo Carver, a developer at SalesRacer. "When you combine that with some kind of an AI engine, then you are talking about something that could really be revolutionary."
Rodolfo Macias is a developer at Sweb Development, which helps businesses sell products or ideas. "Seeing the potential to be able to use their AI system — like a bot — to help them sell, get leads, filter out leads using that technology is going to be amazing," he said.
The keynote sessions focused more on users rather than brands. Facebook is applying artificial intelligence to everything users share — from status updates to cat videos — and revealed some recent advances at F8.
For example, there are more than a billion new stories posted on Facebook every day. An individual user likely sees just 15 to 20 stories that Facebook has identified as likely to be most relevant, said Schroepfer.
The long-term goal is to empower users to interact with the Facebook platform itself, and go from passively consuming content to engaging in a dialogue that helps them find the content that is most interesting to them. This may happen through Facebook's virtual assistant M, bots on Messenger, or image and text-to-voice technology that enables the platform to "talk" to users.
On Tuesday CEO unveiled his vision for Messenger as a platform for people to communicate with bots and is making the company's Wit.ai Bot Engine — the smart system which powers Facebook's virtual assistant M along with human assistance — available to developers.
Brands are giving bots on Facebook Messenger a big thumbs up. The opportunity to extend the relationship they have with users and communicate with them in a more creative way is appealing, said Peter Chau, a product manager for The Weather Company, an IBM Business.
"Building that daily habit with our users is something we look forward to doing," said Chau.
The fact that Facebook is making so much of its AI technology available to developers on and off the Facebook platform is particularly exciting, said Brian Jeans, a developer at Microsoft.
"Developers are able to take advantage of the service, and service infrastructure and all of the AI work that Facebook is doing, contribute to that and it's a greater overall experience for everyone," he said. Dennis Kearny is also a developer at Microsoft in the Xbox division and is interested in how he can apply Facebook's image recognition technology to filter inappropriate content.
"Our product, we have a lot of children using it and we have a lot of user generated content, also," he said. "For us it's very difficult for us to be able to put that in front of children with confidence that it's not adult content or inappropriate and AI will help us solve that problem by categorizing it as adult so we can protect our users from that kinds of stuff," he said.
Here's a rundown of the most cutting edge work Facebook is doing in AI, according to Schroepfer.
Understanding human language
Recreating natural human-like conversation is among tech's biggest challenges, said Schroepfer.
"When you compare it to something like a factual query on search — how far is the moon from the earth — that's easy," he said. "This requires context and understanding."
The company also rolled out an AI-based automatic translation system to translate posts in news feeds. What makes it different from existing software is its ability to interpret and translate quirky regional slang — think Urban Dictionary plus status updates — helping users across the globe better understand each other.
This could help Facebook engage with users in emerging economies, which are seen as key to sustaining Facebook's growth, as user growth in the the U.S. and Europe slows. Right now 50 percent of Facebook's users do not speak English and 800 million people are already seeing translated posts in their news feed every month.
Finding photos on Facebook
Facebook's automatic image classifiers make it easier than ever to find photos showing specific people, places and things posted on its platform. "AI is making the most progress most quickly in computer vision," said Schroepfer.
On top of this, new image segmentation technology enables Facebook to understand images down to an individual pixel. This means Facebook now identify individuals objects in photos. The company is combining this technology with speech so it can tell a users with impaired vision what is shown on screen.
This has obvious utility for brands - if a user is seen holding a Coke or wearing a pair of Nike sneakers, there is a good chance they are a potential future customer. (t is not clear if this technology is being made available to advertisers.)
The future of video
Mark Zuckerberg expects the majority of content people see and share on Facebook to be video within the next five years. With so much video on the platform, users will need better ways to find the content that most interests them, so Facebook is applying the same computer vision techniques to understand and classify videos beyond the use of tags and associated content.
For example, Facebook is applying facial recognition technology to videos, allowing users to search for a person and immediately jump to the exact frame of video where they show up.
With people's attention increasingly pulled in different directions, Facebook needs to find new ways to surface relevant content and keep users engaged. It will be interesting to see how advertisers jump on this.