Acing analytics with AI


Adrian Pennington , October 5th, 2017

The current pace of development in artificial intelligence (AI) is remarkable, achieving milestones in computer vision, speech recognition, and natural language processing that were, until recently, considered many years away. In combination with large-scale data collection and analysis, AI is also increasingly taking on some operational roles in media production and distribution.

However, much of what is branded as AI in media should be understood as the evolution of data analysis. First, the use of unstructured and structured data, coming from multiple sources, combined with configurable data science algorithms, allows for the development of powerful prediction engines that are galvanising the data analytics and business intelligence market. New knowledge graphs and correlations can be found and refined in real time by data scientists. 

“This is where most value can be delivered in the pay-TV industry in the short term, as the players need to drive their business using more relevant data that they often don’t have, as legacy TV systems are not always two-way connected,” explains Simon Trudelle, senior director, product marketing, Nagra. “Social platforms and other external systems can offer extensive sources of data intelligence that can benefit service providers.”

A second dimension to consider is the development of machine learning (ML) predictive algorithms that use a feedback loop to improve the relevance of the prediction engine over time.  Then there are AI/ML platforms that go further and leverage neural networks. These have been proven to outperform human beings for repetitive skill-based tasks, such as speech or image recognition.

AI, ML, and robotic process automation can be seen as an integrated suite of solutions that has already proven successful at dramatically improving back-office and operational-level processes. Specifically, robotic process automation is directly applicable to routine rules-based back-office and operational processes, which currently consume enormous amounts of human resources and time.  

“Automation of routine and rules-based tasks like report generation, account entry, account closure, media uploads, and a wide variety of financial processes, have three immediate and direct benefits,” explains Chris Hodges, managing director, Accenture – communications, media, 
and technology. He prefers to term this “intelligent automation”.

Firstly, he states, automation makes the process go faster. Secondly, it reduces the errors ubiquitous in “swivel-chair” processes, where a person goes back-and-forth between multiple systems with mind-numbing repetition. Thirdly, automating these back-office processes frees up human capital for more creative and “human-level” tasks.

Some specific examples in media include media uploads, reformatting, file renaming, and responding to network alarm interruptions and outages.  All of these are well suited to the application of intelligent automation, according to Hodges. 

“Once processes are automated, they produce an enormous amount of real-time data, which can be analysed for patterns and trends, and allow for pre-emptive process changes or adjustments,” he says. 

“AI can analyse successful video or media projects based on specified criteria. These criteria might include the number of people, shot angle, gender, movement, sound levels, and so on. All of these can be analysed by AI/ML, producing patterns of production for specific criteria. Today, this is done largely through a combination of producers with particular individual personal styles. AI/ML will allow this to become more systematic, repeatable, and scalable across multiple projects at once.”

Machine intelligence provides an important toolkit to business, with applications ranging from insight – how systems are actually being used – to augmentation, which invloves assisting human effort and oversight, and automation, via autonomous system monitoring and intervention. 

Ericsson’s CTO, Steve Plunkett, identifies examples including rethinking the quality control (QC) process by analysing errors found in manual and auto QC activity and identifying higher risk material, based on the supplier, technical properties, and so on,  where more time can be spent, rather than a uniform approach to all content.

Video management software and services company, Piksel, incorporates AI/ML into its technology. It is able to conduct automated inspection of content to provide deeper metadata identification and linkage. “Machine learning can be used to match shows and movies with a greater than 95% accuracy, so service providers can use this improved consistency to give their customers better search and UI provision,” says Kristan Bullett, head of solutions. “Machine learning can also assist with matching to third-party metadata providers to further improve accuracy.”

In terms of cybersecurity and the threat of malicious attack, if the system is monitoring all the logs and knows what ‘normal’ looks like, then unusual internet traffic coming from a new device on the network could be flagged as it happens, enabling early intervention with preventative measures. 

“Rather than just providing intrusion detection in the network layer, our automation application software might become aware of where control commands are expected to come from,” says Pebble Beach Systems’ CTO, Ian Cockett.

Another important area is in network management. Bandwidth is not limitless and is a significant portion of the expense of operating, especially for over-the-top (OTT) content. 

“Today’s CDNs (content delivery networks) and protocols are very wasteful of bandwidth – your device will always pick the highest available bandwidth, whether it’s needed or not at that particular time,” says Tim Child, co-founder at media asset management (MAM) vendor, Cantemo. 

“Analysing the video as it is transmitted and then using the data to determine the required bandwidth can help network operators make more efficient use of bandwidth, lowering costs and improving quality at the same time.”

Limits of training data

To be useful, however, you first need to collect lots of data, know how to use it, and have a purpose or strategy in place. The industry is quite immature when it comes to the use of data-driven insight and processes, but a lot of experimentation now seems to be underway. Netflix and Amazon, for example, already use data extensively to commission original programming and curate ‘discovery’ based on user feedback. 

“Media applications and infrastructure, in general, don’t emit enough useful data, and their control systems and interfaces are designed for people rather than machines,” says Plunkett. “Without enough data, machine learning can’t be effective, and unless we can control systems with machine-friendly interfaces such as APIs (application programming interfaces), then we limit our ability to increase automation and insight.”

Hodges agrees: “The biggest drawbacks today are the lack of existing data to feed the AI engine, and a lack of understanding about how the AI engine can help.”

Rajeev Dutt, CEO and co-founder, Dimensional Mechanics, goes further.  He calls most AI “stupid” because machines are typically trained on narrow pieces of data. “Anything outside of that dataset and the machine can’t understand it,” he says. “A lot of times the project will fail because the training does not encompass broader data sets.”

A neural network trained on a dataset of faces, for example, could have built-in bias because it is seeing only the majority of faces belonging to one ethnic group. 

“You need a greater level of specialisation,” Dutt says. “An AI for Fox News will not necessarily work as accurately with an MSNBC audience, and vice versa.”

Bullett advises that media organisations will need time to adjust to AI/ML, and he expects some organisations to be very cautious at the prospect of machines ‘freeing up’ staff time, as the finance team evaluates cost-saving options in this space. 

“However, ML/AI does not need to be about resource reduction, and can be used to provide powerful tools for operators looking to make better use of data, for example segmentation information or data to improve quality control,” he says.

Media companies are advised not to rush to embrace machine intelligence but to instead spend time learning the basics and discovering how it might help to improve business decisions and performance. “They should then experiment to find practical implementations that work for them,” adds Plunkett. 

“This is best achieved with external help and, depending on the results, they will probably need to hire experts in data science, data engineering, and so on. One word of caution: embracing data-based insights and operations requires a mindset change, ongoing investment, and can be met with significant cultural resistance; simply hiring a few propeller heads into a new data science department will not create a data-driven business.”

Such advanced systems will change current practices, but require supervision and proactive management to make sure key business objectives are reached. Constant benchmarking with other predictive applications is often needed to validate the performance of the new system. 

“It’s clear that while automated AI/ML technology can help reduce the workload associated with some tasks and improve business performance, it does not take away the other more business-centric angles that need to be considered when deploying any IT-based solution: scoping the business issue, defining the ‘why?’ and ‘what if?’ questions, making sure the outputs have an actionable business impact,” says Trudelle.

Looking to 2020

This is one of the most rapidly growing tech trends in the industry. Capabilities – even just three years from now – will far exceed what is currently possible.

“We will see leaps in GPUs (graphics processing units) and CPUs (central processing units) to power cognitive apps that will allow us to train much more complex models and expand the range of things that a neural network can handle,” says Dutt.

Niall Duffy, CMO at Virtual AI, which specialises in robotic process automations, predicts that by 2020 AI will be able to compile schedules and automatically edit or version ‘standardised’ video content (i.e. where editing is more predictable) for updating videos on social media or compiling highlights packages.  “It will also start delivering personalised content in a fine-grained manner, rather than the more blunt-force approaches of today,” he says.

As companies like Microsoft open up their AI libraries and functionality within their cloud products, it allows the tools to be used to automatically analyse content and write or extract metadata. That subsequently enables providers to focus on processing content in a meaningful way, and as AI algorithms get better, more content will be processed by using them.

Cantemo’s Child forecasts that personalised video content will be generated based on audience profiles, behaviours, and locations. “There’s already a great deal of investment from media companies capturing detailed information about viewers to enable them to serve personalised content and ads,” he says. “Using video intelligence, this process can be automated and much more accurate, meaning consumers only get content that is highly relevant and interesting to them.”

Oliver Botti, head of international business development at Fincons Group, agrees, linking AI with the shift of TV away from being used by advertisers as a mere ‘shop window’ and towards becoming an interactive ‘store’. 

“In the near future, moves will be made to further integrate mobile viewing screens and TV in the intelligent home environment, further increasing the volume of potential data served up for analysis,” he says. “As brands start to experiment with broadband/broadcast convergence, editorial models of traditional TV will combine with self-service or recommendation-based models typical of web and mobile, towards new data-driven paradigms: here is where AI will increasingly stand out as a critical tool to help broadcasters tap into innovative advertising, content, and service offerings.”

What artificial intelligence and machine learning are unlikely to solve any time soon are challenges that require empathy, social intelligence, and creative intelligence. However, if today’s mindset is man versus machine, the future – according to consultants PwC – is that man and machine together can be even better than human. 


©2017 ITP Business Publishing Ltd. | Use of this site content constitutes acceptance of our User Policy, Privacy Policy and Terms & Conditions.