Navigating the rise and rise of artificial intelligence in media

How will AI systems inform our decisions?

The chart below was produced by the IEEE back in the year 2000 and was more recently reproduced by KPMG’s Chief Scientist Andrew Morgan at the Microsoft Future Decoded event in London earlier this month.

Morgan said when it came to AI and robotics the chart still represented the best ‘swim lanes’ for how humans should decide who makes decisions.

Listed at number 10 is the highest level of interaction which says: “The computer decides everything, acts autonomously and ignores the human.”

It begs the question, when should computers, even ‘artificially intelligent’ computers make decisions for us? And if that is going to happened shouldn’t we understand a little of how we get there?

In the mainstream press there are many headlines about AI. The predictions of a dystopia where a swarm of AI robots take over the planet and enslave humans are those which get the widest coverage.

The second type, which could be classed as ‘the concerned view’, appear regularly in the business pages and business press. They tend to cover forecasts of how machine learning and automation will drive millions of human jobs from existence in a very short timeline.

Again, the forecasts are that the low-skilled repetitive tasks will disappear first but that ultimately no-one will be spared.

This article is a third type and hopefully a rational view for the media sector.

The media sector, in common with many others, is at the beginning of a disruption. There are clues emerging from use cases which are starting to reveal what that disruption could look like for particular business categories across the sector.

If the forecasts are correct, AI, machine learning and deep learning have implications for broadcast, media and entertainment on a par with the rise of the internet.

Melanie Warrick, Senior Developer Advocate at Google told the MCubed conference in London recently that AI Narrow Intelligence represents the first step into the AI world and for Google is used in products such as Android, Apps, Gmail, Translation and YouTube.

AI has yet to master complex problem solving, emotional and social IQ and rule changes, she says.

So questions for broadcast, media and entertainment professionals fall into several categories.

What can we achieve with AI today? How should media and broadcast seek to engage with AI platforms? And what do I need to know to exploit AI?

We can approach AI Narrow Intelligence through machine learning. Machine learning is a subset of AI based on mathematics and statistics. Given enough computational power and time the algorithms are learning from the data and not from the software code.

A use case for a broadcaster or content owner may be: I want to know how many people watched my advert? How many times did they engage with it? How many were men who are interested in my product?

In June this year, IBM-owned firm The Weather Company said it produced the first cognitive ads for the auto industry. They were available on the Weather Channel app and on weather.com. 

“The benefits of leveraging Watson Ads to not only engage consumers, but also extract valuable insights” - Sarah Ripmaster

The cognitive ad format combines machine learning, natural language understanding, and integrated dialogue tools. Through Watson Ads, Toyota is harnessing the power of AI to engage and educate consumers about Prius Prime - addressing consumer questions, sharing new car information, and guiding decision making during the purchase consideration stage.

“Toyota has an organised system for their data management along with internal processes, which allowed them to recognise and leverage Watson innovation in advertising. They were quick to recognise the benefits of leveraging Watson Ads to not only engage consumers, but also extract valuable insights gained from those interactions,” said Sarah Ripmaster, Head of Automotive Sales, The Weather Company.

Another example may be found in the discipline of Computer Vision: How can a computer learn to recognise and understand images or video?

Some AI and deep learning proponents claim their deep learning systems have rapidly achieved a higher accuracy rate than humans for correctly identifying images. For others this level of capability for a computer to understand complex images remains a long way off.

When it comes to image recognition the simple question with the complex answer in Narrow AI remains: How many images will it take for the computer to see before your AI system can always tell a dog from a cat?

Where to start?

For those without a specialist IT background the discussions around neural networking, data warehousing, big data or data science and onto ML, AI and cognitive rapidly become very technology heavy and very challenging.

These conversations will include which programming language to learn. Python? Perl? PHP? Ruby? Java for front end? Lisp? C++? Which library to choose which is best suited for your requirements. Tensorflow? Theon?

For a walk through of these tools, languages and libraries there are some guides. Github has a list of machine learning frameworks, and software by language here. Less comprehensive lists can be found here and here.

Some libraries are open source and there are a variety of emphases across the tools for different applications such as computer vision, image recognition, speech recognition, natural language processing, speech to text, and data mining to name just a few.

And ultimately the question will be which AI platform to choose? IBM Watson? Microsoft Azure? Google Cloud AI? Amazon AWS? Another platform?

These conversations rapidly become about skills, resource, management and return on investment.

How to engage?

KPMG scientist Andrew Morgan’s view is that high value decisions are always taken by humans: E.g. Should we build that bridge? Should we invest in that movie? While low value decisions are currently the realm of robots.

A simple search will show there are many engagement frameworks for the different aspects of AI.

Is AI new? No. it has been around for 50 years. Is machine learning a new way for computers to operate? Yes. By learning from the data sets, computers are learning about patterns and behaviour and therefore providing more valuable information.

At an enterprise level broadcast, media and entertainment firms are embracing AI for pilot and production projects. Many technology firms who are moving into the content space are already using Narrow AI to improve their products, provide better customer experience through relevant content and inform their investment decisions.

They key is informed engagement. AI adoption is here and much of it will be baked in to other IT solutions. Becoming familiar with the technology is the only way to ensure it is maximised to the full.