Software Infrastructure Global Viewpoint – September 2018

BBC Harnesses AI for Channel Scheduling and Program Creation

The BBC has outlined how it is using techniques under the banner of Artificial Intelligence (AI) to unearth content viewers would like to see from its vast archives.

Then the BBC is taking the same archive-scanning technology one stage further by assembling clips into a completely new program which it is presenting in early September called “Made by machine: When AI met the archive – a new and experimental programme exploring how BBC R&D’s technology works.” This is presented by Hannah Fry and a virtual co-presenter, comprising four segments of archive clips sourced and edited together by the AI algorithms.

This is done by combining four core techniques. First, the system learns to identify what a scene consists of, classifying it by type of landscape and nature of objects within it, as well as whether people are featured and what they are wearing. The second segment scans the subtitles of archive programmes and then looks for connections between words, topics and themes as it pieces footage together. Thirdly the system analyses whether there is a lot of activity on screen, defined as high energy, or not (low energy). Finally, the system draws on what it is learning using all three techniques to create a new piece of content.

The BBC makes the common mistake of overplaying the AI aspect as if this represents an utter revolution in video technology when it is just part of the evolution towards more sophisticated and deeper audio/video search. It is far better just to focus on the capabilities which in this case have great potential for unlocking the power of the huge archive dating back to 1953 and comprising over 250,000 programs.

The real value will lie not in the relatively gimmicky creation of an “AI generated program” but in finding whole episodes or series previously overlooked by the BBC’s human researchers. These can now be “re-released” for today’s audiences. Related to this is the potential for recommending content much more effectively and drawing on this huge reservoir of content, with ability to perform intra-search for clips within whole titles. For a public service broadcaster like the BBC still largely dependent on license fees that almost certainly will not continue forever, this offers hope of a major new source of revenue.

“Helping BBC Four scour the BBC’s vast archives more efficiently is exactly why we’re developing this kind of AI,” said George Wright, Head of Internet Research and Future Services, BBC R&D. “It can find very specific pieces of content in large libraries of programmes, which has massive benefits for BBC programme makers and audiences.”

Commenting is not available in this channel entry.

Related Editorial Content

AI: Real or Hype?

Many new software applications now claim Artificial Intelligence (AI) or machine learning as their underlying power. Is this just marketing hype, or is AI going to be really useful technology for the media and entertainment sector?AI is not new,…

What Separates Artificial Intelligence from Automation?

More so than many other technology trends, artificial intelligence (AI) has quickly become one of the main buzzwords in broadcast and media production. As a result, marketeers in every area of the industry are now presenting products that integrate AI…

Can Content Create Itself?

A revolution in storytelling for TV, cinema, VR, and related forms of entertainment has just begun, enabled by artificial intelligence (AI). This computer-science and engineering-based technique, including machine learning, deep learning, language understanding, computer vision, and big data - is…