Facebook is Getting Serious About Business
I don’t know if you’ve noticed, but Facebook is starting to turn towards a business platform, competing with the likes of Google and Microsoft. The core of this strategy is their “Workplace by Facebook” offering, which was introduced in 2016. They’ve designed Workplace to go toe to toe with Microsoft teams and Google’s G-suite collaboration. In that short year and a half, they’ve seen growth to over 30,000 organizations, with some big names like Starbucks and Virgin Atlantic. At BA Insight, we have noticed this too. Customers are abandoning Yammer and Jive and moving to Workplace by Facebook.
The other thing that is interesting to me about Facebook’s business aspirations is the underpinning of the AI capabilities that are popping up across their stack. Facebook enjoys a data set of raw human behavior that simply does not exist elsewhere. Microsoft is likely a close second, in that they have a model of human “work” behavior, but let’s face it, we’re all on our best behavior at work. Will our work patterns accurately predict our behavior? I’m not so sure they will.
How does this help them regarding AI? There are several ways, which I’ll break down:
Bots Are Here to Stay
Bot frameworks are getting a lot of investment across the various AI providers’ stacks. Microsoft, Google, and Amazon all have offerings in this area. Facebook deserves to be in that conversation now. They have integrated the Bot capability that they have matured in their consumer messenger application, bringing this consumer experience to the business world by pushing it to Workplace. This talks directly to how accurate a bot can imitate a real person. If bots are of interest to you, then read more about what Facebook has to offer here.
They Can be Great at NLP
Facebooks messenger capabilities have introduced built-in NLP. This offers them, and developers integrating with their platform, the ability to automatically detect meaning and intent in every message. Bots built on this framework not only get the words the user types, but also the extra details around the meaning and intent of what they typed. That’s an extremely valuable leg up. It becomes very simple to implement a bot that understands that when a user says “Hi”, “What’s up?”, “How’s it going?” or “Yo”, that the user is sending a “greeting” and the bot should respond in kind. Facebook’s data set of conversations available from their consumer platform provides intelligence that they can apply to the business side, which speaks volumes to the quality of data they are able to provide. In my opinion, this allows them to compete in the same playing field as Microsoft, Google, and Amazon.
They Can Be Great at Image Recognition
Facebook is seeing an unprecedented 85.4% accuracy score on their image recognition capabilities. Again, their platform and access to data allowed them to pull ahead of others in this space. Their unique approach included the hashtag data of images. Not only did they have the image, but they also had the hashtag associated with an image. In their words, this included “dealing with multiple labels per image (since people who add hashtags tend to use more than one), sorting through hashtag synonyms, and balancing the influence of frequent hashtags and rare ones. To make the labels useful for image recognition training, the team trained a large-scale hashtag prediction model.” Of course, they also fed over a billion images to their training model!
They Even Have a Framework
Interestingly enough, while they were rolling out all those great AI features, they needed a framework that could handle their data sets and varied projects circling AI. Out of that work came the deep learning framework they named PyTorch. Once they got it to a point that it did what they needed, they open sourced it for others to leverage. It supports multiple operating systems and a few different package managers, and it comes with some example projects and tutorials to get developers started quickly. It’s seeing usage expand rapidly, and it is already being used by Twitter, SalesForce, Uber, and some big names in higher learning (i.e. Stanford, Oxford, and Carnegie Melon). To me, this is the most exciting outcome of Facebook’s AI push, as the more open source capabilities we can get into the market, the faster these capabilities can innovate.
What Does All this Mean For You?
This is the easy part. AI capabilities are very broad and should be implemented to solve specific problems. Picking the right AI provider and integration is more about aligning what you want in the end, with what the AI provider has a proven record for doing. Facebook has access to patterns of human behavior not available to anyone else. They also have access to more image content than you can imagine. Couple all of that with an open sourced framework that they use internally for all their projects, and it becomes a complete picture. There is value in their experience, and it shows in where they are innovating in AI. They deserve to be part of the conversation.