Artificial intelligence (AI) has been much in the news of late. The bankruptcy law firm of Baker & Hostetler recently hired an “AI lawyer” based on IBM’s Jeopardy!-winning Watson to tackle legal research, the messaging service Slack is training AI bots to be virtual office assistants, and Google’s DeepMind AI recently bested a human in Go, an ancient strategy game that many AI researchers thought would be beyond the reach of software mastery for years, if not decades.
The rapid advance of AI is also radically transforming the digital content economy, even if it’s not provoking apocalyptic hand-wringing from the likes of Elon Musk and Stephen Hawking.
From music, video and photography, artificial intelligence (specifically machine learning) is playing an increasingly critical role in learning about customer habits and serving them up content recommendations.
Netflix famously pays people to watch its videos and tag them—a job that superficially sounds tantalizing until your realize you have to watch movies like The Human Centipede 2 through to the end. But Netflix also retains the services of machine learning/artificial intelligence to help it better understand what its millions of viewers might want to watch next. In fact, it once dangled (and then paid) a $1 million bounty to any software engineer, or team of engineers, who could improve its content recommendation software by just 10 percent.
As video services play host to ever larger libraries of content, retaining eyeballs is increasingly a matter of delivering (and anticipating) what consumers want to see. Given the size of the libraries in question, it’s a task no human can perform. This goes double for user-generated content sites like YouTube, Vimeo and, increasingly, Facebook, that play host to tens of millions of videos. Similarly, broadcasters are leveraging this technology to improve the personalization of their video-on-demand and OTT assets.
An analogous process is underway in digital photography and, in particular, stock licensing agencies that are sitting on millions of image assets that they need to monetize. Companies like ImageBrief and Pond5 have tapped machine learning to understand what customers are searching for and what creatives are uploading, so they can better tag images and stock video assets to surface them in searches. They’ve also taken a more ambitious step beyond merely cataloging to judging an image’s esthetic qualities, to enable sorting by more subjective criteria. (A fascinating breakdown on how one service, EyeEM, did this was recently published on Medium).
The race to implement and improve on machine learning isn’t simply a matter of convenience—it can be a critical means of retaining customers or poaching new ones. Nowhere is this more obvious than in music, where, unlike in video, there is much greater content homogeneity between services like Apple Music, Google Music and Spotify.
Spotify’s “Discover Weekly” playlist—an algorithmically derived weekly selection of music based on a user’s listening habits—has been called “scary good” and is credited with helping the service retain and grow its customer base in the face of stiff competition from Apple, Google and Amazon.
In the new digital economy, it’s not simply about who has the rights to great content, but who can build a system smart enough to anticipate the desires of millions of users. Welcome to the future.