AI and machine learning make inroads in broadcast – but ROI will need to be seen in 2019

Video Net

If 2018 was the year when AI/ML (artificial intelligence and machine learning) took off in video production, then 2019 has to be the year the industry starts to see some payback- otherwise it may go the way of other recent ‘flash in the pan’ tech trends like 3D and VR.

http://www.v-net.tv/2018/11/27/ai-and-machine-learning-make-inroads-in-broadcast-but-roi-will-need-to-be-seen-in-2019/

That is unlikely given that AI/ML – no matter how you slice the definition – is not a niche product but a global tool for vendors to develop media specific applications. According to the IABM’s latest bi-annual Buying Trends survey, which tracks technology trends, actual AI/ML adoption in the broadcast and media industry is up from 2% to 13% in just six months from April to September this year.

The survey data shows that it is larger organisations that are much more likely to deploy AI technology, with adoption varying across different segments of the content supply chain. Another 68% are likely or very likely to deploy it in the next 2-3 years.

While broadcasters are undoubtedly experimenting with AI/ML tools to automate and augment current workflows, as one unnamed broadcaster told the IABM, “The challenge is monetising it, or creating real commercial value”. In other words, the use of AI/ML demands a real business case, otherwise costs can quickly escalate to outweigh any potential advantage.

Alongside the cost/benefit equation, there are also other challenges that may prevent media companies from taking full advantage of AI/ML. Namely, it is best suited to working with large amounts of data.

“While subscription-based broadcasters and media companies already have significant viewer data to work with, FTA [free-to-air] broadcasters who are moving into OTT need to build new data to better understand their customers; this is why the uptake is slower than it might be,” infers IABM CEO Peter White. “Other challenges relate to data management [i.e. training, and updating the data] and data gathering. Also, media companies need to manage different types of data together in a single pool – avoiding data silos – to create real value from AI/ML algorithms.”

IABM expects AI/ML adoption to increase in 2019 as current solutions mature and broadcasters build their databases, enabling them to drive more automation and liberate resources within their organisations. There is definitively a growing interest in the many potential applications of AI/ML, but this is tempered by an overriding need – as with any new technology – to prove its business value.

The trade body, which represents the interests of equipment manufacturers, also charted a slight rise in adoption of IP technology in the production sector this year. An example of this trend is UHD production: IABM data shows that companies that intend to move to UHD are much more likely to invest in IP technology for its format independence.

IABM does observe that most broadcasters have a cautious culture that is holding back transition. “Many broadcasters have been reluctant to abandon SDI and have so far been inclined to adopt a ‘wait and see’ approach,” notes White. “Maybe the floodgates are now beginning to open.”

The finalisation of the SMPTE transport protocol ST 2110 earlier this year has given IP roll-out an uptick. But this is a slow burn. As the IABM underlines, “IP not only impacts technical facilities and workflows, but also the business and cultural environment.”

From a cultural perspective, there is also a convergence between IP and broadcast that is very difficult to put into practice at traditional organisations. The move to IP is as much about transforming culture as technology change.