How Facebook is using machine learning
The awesome computing power of Facebook can’t be sustained by engineers alone – and so the social network is developing more and more processes that use machine learning (a form of A.I.).
In 2011, the big problem facing the company was how to build the infrastructure needed to maintain the site without a prohibitive price tag. By taking the problem in-house with the Open Compute initiative, Facebook succeeded in building a sustainable and dramatically cheaper solution.
The problems five years later stem from how information is categorised and presented to the 1.5 billion users of the site. Tasks such as tagging photos for the visually impaired, filtering messages for spam or translating posts in your newsfeed can be done by Facebook’s internal machine learning platform FBLearner Flow – leaving engineers and editors free to develop even more imaginative features.
Inside Facebook’s biggest artificial intelligence project ever
Fortune 14 mins
Facebook has grown up – and taken over the world
30% of the time Americans spend on mobile is spent with Facebook and other apps owned by the social media giant.
The sages of Silicon Valley used to chuckle at Mark Zuckerberg’s ten and twenty year business plans – the accepted wisdom was that social networks never lasted. But the acquisitions of companies, including Instagram, WhatsApp and Oculus, has extended Facebook’s reach, ensured its dominance and expanded the huge amount of data it has to leverage ad spend.
Zuckerberg’s vision to become a ‘utility’ is now within reach – many of us use Facebook as a single-login for huge swathes of the Internet. If the company succeeds in becoming completely indispensable to users, it will remain that way for advertisers.
How to win friends and influence people
The Economist 18 mins
Ad revenue everywhere else is drying up
According to Morgan Stanley, 85 cents of every dollar spent on digital advertising goes to Facebook or Google – leaving news sites and publishers in a precarious position.
Audiences are preferring to access content from within their social feeds and many publishers are seeing the corresponding drop in traffic. The options for monetising editorial content are now to either join the Facebook family using Instant Articles and paid posts or develop partnerships with ecommerce sites (dominated by Amazon) which generate revenue for purchases of featured products.
Alternatively publishers are partnering with brands to create content that appeals to their combined audiences – as BuzzFeed have successfully demonstrated.
Media websites battle faltering ad revenue and traffic
The New York Times 6 mins
Connecting the dots
We’re big fans of the truism that innovation happens when disparate dots are connected (as you can tell from our annual conference) and this article is a great reminder.
When we’re spending more and more time on social networks that reflect our own ideas and communities back to us, it becomes even more important to seek out other ideas and perspectives. Pay attention to the world around you, visit new places, museums outside of your specialism, conferences aimed at other professions, read books that you would normally pass by.
The secret to creativity: become an intellectual middleman
Fast Company 4 mins
Taking risks to transform communication within your organisation
Brands know that communication with their customers is now a two way conversation – it is time for internal communications to catch up. In this post Culture Director Belinda Gannaway takes us through the six boldest characters in internal comms; instigators, data junkies, risk takers, silo busters, outsiders and those who are totally people-centric.
What does it take to transform internal comms?
Brilliant Noise 2 mins