In 2017, artificial intelligence (AI) deep learning firmly entered the media and entertainment lexicon. While the idea of AI winning games of Go may grab the headlines, advances in AI and deep learning are also helping media providers to gain -- and act upon -- greater insights from their ever-growing mounds of data from connected devices.

This is significant at a time when TV viewing audiences are fragmenting and competition from the likes of Netflix, Facebook, Google, Amazon and Hulu is intensifying. The end game is to apply AI and deep learning techniques to drive efficiencies across the business, offer a range of personalized services and open up new revenue streams.

Interestingly, though, AI and deep learning have a certain Back to the Future ring about them. In 2006, when our first customer went live with a personalized content recommendations engine built on our deep learning platform, it was called "data mining." But whichever name is the flavor of the moment, the mix of machine learning algorithms and techniques has been evolving ever since.

Today's personalized recommendations now rely on a mix of AI and machine-learning algorithms to make sense of volumes of content and metadata across multiple platforms and devices, including user preferences.

What's fuelling this AI boom are two technological advancements. First, AI users no longer need experience in statistical analysis or a PhD to take advantage of the technology. And second, the elastic computing power available in the cloud makes it faster and cheaper to build and scale new applications.

Facebook, Amazon, Netflix and Google have all exploded because they set out to understand their customers' behavior right from day one. While OTT video and pay-TV operators have done a great job amassing data lakes about viewing behavior, many have not yet taken advantage of AI and machine learning to raise their game in a competitive marketplace.

As AI and machine-learning algorithms and techniques sweep through the industry over the next few years, new tools and applications will emerge. Here are some examples:

Minority Report comes to the small screen with predictive recommendations
Capturing and managing TV/video platform data from multiple sources for analysis using advanced, predictive algorithms is a key focus for the media industry. By using AI and machine-learning predictive algorithms to understand what is happening across the entertainment platform -- for example, what consumers are doing on each device, and what type of entertainment content they are engaging with (e.g., TV, video, music, games) -- you can predict what they want to do next and serve up the right content at the right time. Adding a feedback loop allows you to improve the relevance and success of the predictions engine over time.

Analyze this, then act
AI and deep learning are also at the heart of a new genre of big data analytics platforms aimed squarely at the TV space. Some innovative pay-TV operators are already analyzing data in real time to realize real business value and enhance the user experience.

These insights solve specific problems -- such as how to boost engagement, stop subscribers from cancelling their subscriptions or entice them back to your service. They can also drive decisions on content buying, pricing, marketing campaigns and editorial promotions. And soon we'll see this data used as a basis for new revenue-generating opportunities, such as dynamic ad insertion.

Lost content and the need for universal search
Right now most US consumers subscribe to four TV services, according to PwC's Consumer Intelligence Video Service 2017. And they are forced to search and navigate each one individually. This is unsustainable. As a result, the industry is moving towards universal search, recommendations and navigation. This is possible today from a technical perspective, but we need a level of convergence between services to deliver the reality.

"Talkin' Loud and Sayin' Something" (with apologies to James Brown)
Voice applications are becoming more popular in the home. AI is used both to determine the intent of the voice request and to pass that intent to the right device -- for example, the TV, not the game console -- as well as to handle the more specific intent once the right device has been activated. Right now, on most services, it may only be possible to conduct voice searches on, say, actors' names or movie titles. But the possibilities are truly endless.

As with universal search, consumers need to navigate around the content on all TV platforms, gaming apps and radio channels using a single voice-input device. So service and content providers will need to work together to offer an integrated experience to the consumer.

And, of course, it’s not just entertainment gadgets that are getting the voice treatment. Today AI is a big part of the Internet of Things (IoT) and we are becoming used to voice-activated digital assistants, light switches and thermostats. Soon we will be in danger of waking up ten devices every time we say something in our smart homes!

To combat this, we need to see a convergence around voice-based command and control of the home. There will need to be discussions around who “owns” voice in the home -- the pay-TV provider with the voice-activated remote or Google and Amazon with their respective Home and Echo devices.

More than a house of cards
It’s clear that we've just begun a fascinating AI and deep learning journey in the entertainment space. A good place to start is to use AI and deep learning to get to know your customers better -- to keep them loyal, boost engagement and ultimately drive up revenues. This is what Netflix does so well.

Fortunately, the latest personalized recommendations and big data analytics can be integrated with existing pay-TV and OTT video providers' platforms, delivering them with the customer intelligence and insight required to match, and even beat, Netflix at its own game.

— Peter Docherty, Founder and CTO, ThinkAnalytics