Tech Features

2018 is the year of machine learning: Glodina Lostanlen of Imagine Communications

Glodina Lostanlen, CMO, Imagine Communications.

New artificial intelligence capabilities will dig deeper into the decision-making realm freeing the industry to create engaging content, writes Glodina Lostanlen of Imagine Communications.

The media industry has traditionally been technology-driven. We still see this. Recently the talk has all been around IP connectivity. In truth, this is not a goal; it is not a technology that is interesting and important in itself. It is an enabler — it allows us to achieve exciting new things.

It looks like 2018 is set to be the year of machine learning, or what many call artificial intelligence – AI. But again, we have to be careful not to get too obsessed with technology for its own sake. We need to look at what AI can enable.

The primary aim of the media industry is to create interesting and engaging content, deliver it to consumers and earn a fair return on that content. If technology can help us do any part of that better – make more engaging programmes, increase revenues, connect in new ways – then it is worth pursuing.

The core challenge facing the whole of the industry is that competition for revenues is fiercer than ever. But the cost of the raw material – programmes – is not going down. Indeed, demand for high-quality content is such that costs are soaring.

The first television drama to command a budget of USD1m an episode was Hill Street Blues in 1986. Given the rate of inflation since then, premium productions should be costing around $2.25m an hour. But according to industry sources, popular programmes like Game of Thrones have production budgets of around $12m; period dramas like The Crown $10m.

“Machine learning is used to process social media responses to content … smart algorithms can readily tell genuine criticism from irony, to give genuine insight on audience reaction” Glodina Lostanlen, CMO, Imagine Communications

Media companies are stretched, trying to sustain revenues at more or less constant levels, while paying more for new content. The only solution is to use technology to reduce operational costs.

The much-hyped move to IP is in fact an enabler for the move to software-defined infrastructures. By implementing virtually all broadcast hardware as software applications running on standard computer hardware, the cost model is transformed.

Traditional big box devices become virtualised software. Typically, these will run as microservices, small packets of code to do one precisely targeted function, and which can be combined in different ways to achieve many different tasks. A video noise reduction microservice, for example, could be part of an encoder at one moment and part of a distribution amplifier at the next. Any workflow can seamlessly call any microservice, from any source, to achieve precisely the required functionality.

I am excited about AI because it fits perfectly into this microservices architecture. It can be part of the orchestration layer, or machine decision-making could be a microservice itself, determining how other services function.

This is not a vision for the future. This is happening right now. By introducing machine learning into the content data flow, new and unique functionality can be added. This leads directly to a reduction in operating costs and the reassignment of personnel to more creative tasks, the end that everyone is seeking.

AI and big data are already familiar in many media applications. Maximising audiences means knowing them, and big data systems which support subscriber management typically use AI to analyse trends and demographics.

AI engines provide automated support for content logging, sifting through video and audio to create precise, rich metadata. This can then be used to drive recommendation engines, providing detailed information to sell a programme to an audience. Rich metadata also means more precise archiving and cataloguing, providing media companies with the ability to instantly locate sought-after footage for various production outputs.

Machine learning is used to process social media responses to content, either for direct display on screen or to develop trend analysis to see what is working and what is not. Smart algorithms can readily tell genuine criticism from irony, to give genuine insight on audience reaction.

What we will see in the coming months, though, is a new wave of AI. This will bring with it new capabilities that will dig deeper into the decision-making realm currently solely inhabited by humans. We are likely to see machines take a much bigger role in the creation of trailers and marketing campaigns, for example, particularly to eliminate the tedious task of creating all the different versions of a promo.

While it is unlikely that a machine will ever replace the creative team on a high-concept show like Game of Thrones, in 2017 the BBC ran an experiment which used an AI system rather than a director on a comedy panel show. It first analysed thousands of similar shows, before making a reasonable job of directing what is, to be fair, a fairly formulaic genre.

“We aren’t trying to replace TV directors,” Matthew Postgate, BBC CTO, is reported as saying. “But because we don’t have infinite resources, this could open up the number of events we can cover.”

Some may protest that this is stepping too far over the line between machine learning and human creativity. But there are many situations where decisions need to be made on the fly, often at any time of day or night, where it is reasonable to use machine learning to provide reliable and fast reactive management.

With the need to maximise revenues, another area where AI is already being implemented is in advertising campaign management. The next-generation advertising management platform brings campaign planning across broadcast and on-demand platforms into a single system, allowing buyers to specify precisely the consumers they need to reach.

Without machine learning, providing that defined reach accurately but with minimum use of inventory would be a hugely complex task. If it could be done manually, it would require so many man hours that any savings would be more than negated in additional costs.

As we are proving with media companies worldwide, machine learning can allow accurately targeted campaigns to be created and implemented automatically.

That is where AI can provide real benefits to our industry in the very near future. The media business is very much about people: it relies on creative talent and an instinct for connecting audiences. Machine learning systems should augment what they do. By eliminating manual and repetitive tasks, they should allow the people to be even better at what they do.

As Manjunath Bhat, Research Director at Gartner, says: “Robots are not here to take away our jobs. They are here to give us a promotion – I think that is the way we should start looking at artificial intelligence.