If you believe anything can and will be automated with artificial intelligence (AI), then you might not be surprised to know how many notable media organisations including The New York Times, Associated Press, Reuters, Washington Post, and Yahoo! Sports already use AI to generate content. The Press Association, for example, can now produce 30,000 local news storeys a month using AI. You might think that these are formulaic who, what, where and when storeys and you are right, some of them certainly are. But, today, AI-written content has expanded beyond formulaic writing to more creative writing endeavours such as poetry and novels. 

How does AI write content?

The software process that automatically creates a written narrative from data is called natural language generation (NLG). It’s already used for a variety of content generation needs in our world including business intelligence dashboards, business data reports, personalised email, and in-app messaging communication, client financial portfolio updates and more.

The first step in NLG is to define what format of content is desired. Each content type from social media posts to financial reports to poetry has a unique writing style and structure. The narrative design, also known as the template or narrative type, is constructed by the end-user, the NLG solution or by the software provider. Some of the NLG tools available include Quill from Narrative Science, Amazon’s Polly, Wordsmith from Automated Insights, and Google’s Text-to-Speech while some organisations have created in-house tools such as Heliograf at the Washington Post.

As with any artificial intelligence solution, ownership and access to data are crucial. In the case of NLG, structured data is fed into the software and is processed through the “conditional logic” that’s part of the narrative design. The goal is for the output to sound like a human-generated each piece of content.

Why are organisations investing in natural language generation?

As with other implementations of AI, natural language generation allows organisations to process large datasets and create more efficiently than humans can. Organisations who have implemented an NLG solution can produce thousands of more narratives in a sliver of the time it would take humans to write each one individually. 

In addition, NLG enables complex personalization at scale. This can have significant service and overall experience benefits for customers. If your company has a workplace savings scheme, think of the 401K portfolio summaries you receive quarterly. These are likely generated by NLG, but it is highly personalised, speaks directly to you and uses your unique set of information. 

Natural language processing can also make data more insightful and easier to understand for humans who are not data experts. While charts and graphs are visually appealing, it may be a challenge for some people, especially those who aren’t used to analysing data, to extract the important message they should receive from the visualisation. NLG can further engage the reader who is looking at the information with written summaries and key insights to accompany the charts.  

Examples of NLG used today

To cut research time and costs, German bank Commerzbank is using artificial intelligence to create equity research reports. This process is not yet completely automated, but the technology is already able to perform about 75 percent of what a human equity analyst would have done.

There are several automated journalism applications being used in newsrooms around the globe from speeding up research to fact-checking, comment monitoring, streamlining workflow, eliminating fake news and, yes, even writing content. The Associated Press is using AI to write thousands of sports reports. Following the narrative design, the software can scan the data and determine the insights from a game that are important for the reader to know. Knowing the lingo is important for the content to sound natural. The Washington Post uses its in-house NLG tool to create news articles and social media posts.

Many financial institutions are churning out 10 – to 15-page financial reports in an instant by using Narrative Science’s NLG platform Quill. It also creates content for Groupon, Forbes, USAA and more.

Artificial intelligence helps create text summarizations, short and coherent versions of longer documents. This requires the algorithm to understand the source document and then distil the meaning and important details in a fluent summary.

Natural language generation has now been used to create an almost award-winning novel, The Day a Computer Writes a Novel, as well as a Jack Kerouac-inspired narration of a road trip and poetry. While there are still some glitches to work out before there will be full confidence that machines can actually write with the same creativity and ingenuity as humans, these developments can certainly make you ponder what creative endeavours will be uniquely human and what constitutes quality writing.

Machine or human?

Do you think you will be able to identify content that was created by a bot? Try to guess if a bot wrote this poetry. You can also take this quiz from the New York Times to see if you can determine if you’re reading content written by a human or a bot.

Ready to try AI content for yourself? All you need to do is give this AI writer a headline, and it will do all the research for you!