Are algorithms going to replace writers?

“This (Wriber) might replace me.”

– One-third of all writers I meet

I’m an entrepreneur, marketer, and software developer that helps B2B marketers write better content through the use of machine learning algorithms. Scrutiny comes with the territory of operating an artificial intelligence startup. There’s an assumption that you’re going to run hard-working people out of jobs. Groups that have seen cuts across their industry are the most critical.

This man vs machine debate has been heating up over the last few years, even through natural language generators have been around for a couple decades. Article spinners initially set the bar low for automated content because the quality sucked. Nobody was worried.

We’re focusing on this subject now because algorithms are getting better at writing. Human-generated and machine-generated content can be hard to distinguish from each other in many cases.

Both the Associated Press and Forbes generate business reports from automated algorithms. They can reach in financial data from annual reports and quarterly results to produce content:

“Apple’s holiday earnings for 2014 were record shattering. The company earned an $18 billion profit on $74.6 billion in revenue. That profit was more than any company had ever earned in history.”

– An algorithm wrote that

Similarly, GameChanger produces automated content as part of their sports coverage. They can input data from a score sheet along with other historical data about the players, teams, and league to produce an article about a game:

“Benner had a good game at plate for Hamilton A’s-Forcini. Benner went 2-3, drove in one and scored one run. Benner singled in the third inning and doubled in the fifth inning.”

– An algorithm wrote that too

Most writers are quick to point out these examples are very formulaic. There are common interest points found in almost every business report, such as earnings and milestones. The same goes for sports articles. If a natural language generation algorithm was given a solid amount of common data points, it’s not too hard to imagine that there could be some automation used to produce content.

Can algorithms have a human touch?

An algorithm certainly helped me write that header; do you consider my writing to be real? To the surprise of many, there are algorithms that can write fictional novels:

“It is now right at 21:04 in Campo Grande. In some ordinary yet adequate abode an old woman named Abeba, who is rather large, reads the warning message on an over-the-counter drug container. She zones completely out.
It is now exactly 12:05 in Funafuti. In some typical location a person named Kenny, who is of completely average stature, reads a well-preserved note. He looks away, then back.
It is now almost 19:06 in Cayman. In some homey yet run-down residence an individual who is called Hamza, who usually turns to look up to other people, reads the ingredient list on a box of breakfast cereal. He suddenly collapses.”

– Snippet from Nick Montfort’s World Clock, an automated novel

You can see obvious patterns in the language of the paragraphs above. The story is incoherent and you can tell it was written by a computer. Poetry is another creative domain in writing:

“When I in dreams behold thy fairest shade
Whose shade in dreams doth wake the sleeping morn
The daytime shadow of my love betray’d
Lends hideous night to dreaming’s faded form.”

– Snippet generated by SwiftKey

SwiftKey is a popular mobile on-screen keyboard app with prediction technology. Someone spent the time writing Shakespeare sonnets in SwiftKey and was able to write small snippets with simple clicks. It’s cool but useless.

Some content can’t be completely automated.

The key word here is “completely”. Algorithms are far from writing entire Shakespeare plays on their own, and always require a level of human input. Only very short snippets of meaningful content can be produced at this point.

The long-form assets found in content marketing campaigns (blogs, white papers, case studies) would have tremendous difficulty in being 100% automated. The customization and input required for each individual asset would be substantial to produce something meaningful. You’d be better off writing it yourself.

Imagine you’re writing an opinion piece on a subject (e.g., Was Steve Jobs a good person?). How would an algorithm know what to say for you without your input? It wouldn’t! There is subjectivity and multiple angles found in content marketing that changes from company-to-company and person-to-person. Content marketing assets are very different than the business and sports examples based off specific data.

Semi-automated content will make the difference.

Algorithms won’t replace writers but should still be leveraged to write better content faster, regardless of the subject matter. Have you ever researched a topic? You’re using Google’s artificial intelligence algorithms to find those pages.

We can build algorithms that help novices put pen to paper and turn pros into superheroes. They can provide information around specific subjects, prompt us with meaningful suggestions to include in our content, and check for style consistency, all of which let you focus more on writing a great story.