Automated Journalism

Hugh Westbrook, Senior Product Manager, Sky

With Fake News becoming an ever-growing problem, and trust in the media being eroded as a result, it is perfectly understandable that journalists have some fear and scepticism when the subject of Artificial Intelligence is raised. After all, AI has the capacity to generate the kind of fake news which is the bane of the industry, while the technology’s growing ability to create valuable content leaves many journalists feeling vulnerable about their jobs.

As a former journalist who now works in product development, I can fully understand this view. I remember technologists imposing solutions on teams in the past, often with a mandate to change working practices to support a new product or system. These changes tended to land badly as people found ways to resist them. Attempts to introduce change today still seem to be beset by this reluctance to embrace the new, and so AI-driven transformation in newsrooms is slow.

Nevertheless, AI offers enormous opportunities to the journalism industry, in two distinct ways – the automation of tasks which would otherwise be done manually, and the ability to create content which would not be possible without the assistance of AI.

Automating manual tasks frees up people’s time to do more exciting and valuable work. Automated transcription is one area which can benefit journalists. It can process in minutes a task which might otherwise take hours and dramatically speeds up the writing of articles or video subtitles or aids with production and research. AI continuously improves this process by learning specific terminology pertinent to individual users and by improving its understanding of accents, such as foreign speakers of English, a particular issue with the Premier League, for example.

The use of Computer Vision, specifically object recognition, has significant benefits in the areas of research and archiving.

AI is not a tool to do away with journalism, it is an enabler to create better journalism

Currently, large image or video archives can be hard to search when looking for specific items. Computer vision is able to find and log all manner of things which appear on the screen and then make them searchable, meaning that if someone is looking for a particular type of dog, or a specific time that a celebrity appeared in a particular location, all of that information can be stored and then found swiftly

This ability to search and find information is one of the bedrocks behind the other way that AI can help journalism. Because AI can search and catalogue vast amounts of content, in multiple formats, the ability to then search for the unknown, to hunt for connections between an apparently disparate piece of information, can allow new facts to surface or new insights to be made.

It can also allow for new types of storytelling. At Sky News, we ran a project which used real-time facial recognition to identify guests at the Royal Wedding of the Duke and Duchess of Sussex, an undertaking which would have proved impossible without the technology but which proved a hit with the hundreds of thousands of users who accessed it.

AI will also increasingly introduce new ways for people to access the content. As the use of Smart Speakers increase, so the kind of content which people need to produce for these devices will need to become more sophisticated. Currently, News consumption is largely bulletin-based. Much of this is because the devices themselves are not truly conversational – users end up being trained how to ask for things, rather than simply asking. But in time, this balance is likely to shift, and the challenge for journalists is what kind of content they start to produce for smart speakers and whether AI-generated metadata can help to create atomised chunks of content. Will bulletin formats largely remain, or will they begin to deliver content in bite-size pieces which can be returned as answers to specific questions, allowing people to have more of a conversation with their device about the news?

One other area which has caused concern is the ability of AI to produce content consumed by users. Automated stories are now beginning to appear, created by running well-structured data, such as sports or business data, through a sophisticated template. Many of these stories would not have been written at all without automated interference, so this is about creating new content rather than doing away with the old.

This also fuels concerns about the creation of fake news by using AI, but AI has as much power to detect and do away with fake news as it does to create it, and it will become increasingly important for newsrooms to adopt such tools in order to ensure the veracity of the material coming into them.

Because when all is said and done, AI is not a tool to do away with journalism, it is an enabler to create better journalism. It is about a partnership between humans and technology, working together to create accurate and better stories. Journalism is too important to leave it to technology. But it is also too important to ignore the benefits that technology brings.