Algorithms go mainstream as linear broadcasting strives to merge with digital

It was in April 2019 when ITU recognized the role artificial intelligence (AI) systems were playing in different facets of metadata creation, content access, and subtitling from archives, targeting content delivery and workflow optimizations across production and delivery. The recognition came in the form of its recommendation ITU-R BT.2447-0 (artificial intelligence systems for program production and exchange).

However, much before this, the use of AI and machine learning (ML) had made giant strides, which could not have been missed. In January 2018, BitMovin demonstrated AI-enabled algorithmic compression, which could provide streaming rates that were a fraction of what was needed by traditional encoders by analyzing and learning about the nature of content and receivers in an iterative mode. This is critical as broadcasters move to 4K or 8K technologies or to deliver video over mobile networks with need for adaptive deliveries. The need for all video content to be FCC compliant prior to on-air broadcast had been demonstrated by TVU networks in automatically close captioning and transcribing content. In September 2019, they demonstrated that their solution could access all media asset management (MAM) platforms including ENPS, Dalet, and Primestream.

In news dissemination analysis of diverse sources including social media, Twitter, and massively spread news source, chatter plays an important role. Since 2016, Fuji Television has been using a proprietary social media network analysis system (Spectee) to highlight topics that need to be targeted. NHK has been using a neural learning network to beef up its ability and speed to bring the most relevant news.
This AI-enabled world is being driven by new-generation companies. For example, Connekt, the AI-driven technology company, had partnered with Verrance, USA, for ATSC transmissions across all devices – cable, OTT, and mobile – via tools which allow deployment of on-screen experiences for content, brands, and advertisers.

The AI capabilities are indeed being enabled by AI chipsets, developed amongst others by AMD, Google, Intel, Mythic, Baidu, Qualcomm, Graphcore, and the like – a USD 6.6 billion market in 2018 expected to reach USD 91 billion in the next 5 years. This is set to change the entire workflow in traditional technologies such as linear broadcasting. It is today possible to create a deck of cards – each representing an area where AI/ML can provide solutions. These can include automated commentary, foreign language translations, sign language generation, automatic colorization and restoration of content, automatic content generation, transcription, object detection including face or text in scenes, dynamic product placement, speech recognition, and more. But these are all disparate methods and most of them proprietary.

While use of AI/ML, using proprietary methods of integration with linear broadcast systems, is a way of life today, the question is whether this could be standardized across commercially available systems used in production, broadcast, and distribution. AI/ML is growing at a furious pace and creating algorithms which can be highly disruptive to traditional ways of doing business. This, however, needs a standardized way in which traditional process can hook into the AI world.

But the green shoots are now visible as these areas move toward standardization. In September 2018, Alliance for Telecommunications Industry Solutions (ATIS, a body accredited by ANSI in USA) brought out a report on Evolution to an AI-enabled Network. This was significant as the body works closely with the 3rd Generation Partnership Project (3GPP), responsible for all mobile standards including 5G and IoT. This sets out the framework of how an AI-enabled network can be conceptualized and managed. Across the Atlantic, the European standards body ETSI has now a group – Experiential Networked Intelligence Industry Specification Group (ENI ISG) – defining a standardized cognitive network.

The AI-enabled revolution is similar to one leashed by the internet two decades back, which delivered a world unrecognizable by previous visionaries. The new world, which AI is on course to deliver now, would be akin to a fourth dimension – unknown and unfathomed today.