Alt-text: A robotic hand solves a puzzle

Strengths and Limitations of Content Writing AI

By Alexandra Kaste

How artificial intelligence can speed up content creation by generating articles (and why you can't totally outsource writing to the robots just yet).


AI content writing assistant tools are getting increasingly popular in the workplace. While digital agencies use them to generate all kinds of content for their clients, some companies use them when they need to generate content for a specific topic or niche.

The use of AI for the creation and distribution of content is not a new idea. Some experts believe that AI writers will be able to create content faster than humans, while others are worried that it will lead to a loss of jobs for human copywriters.

No, copywriters are not obsolete. AI writers just help them in doing their job more efficiently. They can focus on creativity and emotions while the AI writer takes care of the technical part.

What no one denies though, is the fact that AI writers can do a better job at pulling accurate data from sources and summarizing it, as opposed to human copywriters.

AI writing assistants are not a replacement of human copywriters. They just provide assistance to the human writers by getting rid of writer’s block and generating content ideas at scale.

You Got Me

Okay, so that introduction was an experiment. The entire intro to this article—title aside—was generated using Rytr, a self-described “AI-powered writing tool.” The program was free and browser-based, with options to upgrade later.

The process was straightforward: After watching a brief introductory video, I entered the language, the tone, what the test was for, and my topic described in 100 characters or less.

In this case, English, “candid” and “funny,” Blog writing section, and the title of this piece along with the question “Are copywriters obsolete?”

Rytr then generated four texts for me. Full disclosure, I did mix and match a little bit, choosing my favorite parts and arranging them, but the text itself is almost entirely in its original form. The result of this process is a perfectly serviceable if vanilla introduction that Rytr generated in mere seconds.

So, what does this mean?

I won't deny that an AI comforting me that it can't take my job yet was…unnerving. But in the fast-moving business of content marketing, you can't afford to be a Luddite. So brands, marketers, and copywriters alike need to explore how best to use this tool to enter the future of content marketing.

In the past few years, there has been an explosion of content generators promising to do things better, faster, and cheaper than ever before. But can do they live up to those promises?

How did we get here?

These copywriting programs didn't spring out of the air. Instead, they're using the same system named GPT-3.. Developed by OpenAI API, GPT-3 is the third in the GPT-n series released in beta form in June 2020.

It’s an incredibly versatile system that works from any written prompt that the user gives it. The program works in natural language processing (NLP), a field concerned with computers' understanding and generation of human language.

OpenAI has over 300 apps using GPT-3, all with different emphases. For example, some use the program to identify themes in customer feedback, while others create highly effective chatbots.

Before its release, the most advanced NLP was Microsoft’s Turing NLG that had 17 billion parameters. GPT-3 has ten times as many.

So what does it do?

According to OpenAI, GPT-3 takes text prompts in phrase or sentence form and forms a complete thought in "natural" language. Users can write whatever they want in the program, and GPT-3 will complete the text. You can “program” it by giving it a few example sentences to give it a better idea of the style of what you want it to say.

This simple description underplays the incredibly complex programming necessary for any approximation of human speech.

Language is complex down to its most basic units. Beyond the literal definitions of words, there's common usage, tone, and connotation. So even semantics isn't as simple as you'd think when you realize how many different definitions most words have. For example, according to the Guinness Book of World Records, the verb 'set' has 430 distinct meanings and 'run' 645.

Language production is complex even for people. When the human brain suffers from a neurological illness, speech and language production often are severely affected.

With all that in mind—and knowing computers are most comfortable in 1s and 0s—it's frankly incredible that a program can create writing that bears even the slightest resemblance to natural speech.

What does this mean for content marketing?

While the program's science is fascinating, as marketers, the question we need to be asking ourselves is how do we harness it? AI content creation is still in its infant stages, and new and effective ways to harness the software are constantly being discovered.

It's especially exciting for marketers looking to automate content creation. However, as with any labor and time-saving innovation, it's essential to acknowledge both the good and the bad to prevent cheapening your brand by misusing bleeding-edge technology.

Limitations

With how simple and easy the process is, it can be highly tempting to dash down a few lines in a content generator and hit publish on the result—especially facing a deadline.

Letting an AI do all your copywriting with no oversight can lead to what I've dubbed the Google Translate Debate. If you don’t speak a language, Google translate is going to create something far superior to whatever you’d come up with, and it’ll do it in under a minute.

BUT, a native speaker of a language will almost immediately identify that it was translated by a machine and notice every spot where it dropped the ball.

The same thing is true for AI articles. Someone who can't write or is facing a deadline will get something way better than what they could produce from an AI article writer. But that doesn't mean it'll be good. Humans will almost always be able to spot the difference, creating an uncanny valley effect.

Yejin Choi, a computer scientist specializing in AI, describes GPT-3 as "essentially a mouth without a brain."

Quality language divides amateur brands from professional ones.

In the example I used at the beginning of the piece, I chose the sections that were, for the most part, coherent and logical. While trying to keep the bulk of writing untouched, I still had to do some editing, fixing nonsequiturs and phrases that just felt… off.

Tone is another aspect of writing that is extremely difficult for a computer to master, and it's better at some than others. When I asked for a candid tone, the content generator did a reasonably good job. However, calling anything it produced “funny” would have been a stretch.

If you’ve ever worked with any program that corrects grammar or otherwise seeks to improve the writing process, you know how AI can sometimes widely miss the mark. Any correction software worth its salt comes with an override feature.

Brands that plunge into AI content generation because of pricing will inevitably come across these roadblocks. Without proper planning and a human writer’s oversight, their content quality will take a hit.

Making AI Work for You

Limitations aside, it would be ridiculous to throw the baby out with the bathwater. Just because AI doesn't replace copywriters and editors doesn't mean that it's not an extremely useful tool when wielded appropriately.

We’ve already written on using artificial intelligence to help generate blog ideas. These programs can allow you to extrapolate their concepts rather than wasting brainpower generating ideas en masse.

Computers can free up humans from grunt work, and AI content creation is no exception. As a writer, first drafts take up an inordinate amount of time in the writing process. Just getting something on the page can take up the most time in the whole process. The concept of the bad first draft is a widespread practice to speed things up. With GPT-3, your bad first draft is almost instantaneous.

Andrew Brown, a poet, who experimented with the program, wrote that “GPT-3 is good enough that it produces something ‘worth editing’ more than half the time.” Using this method, copywriters can spend more time in the editing phase, resulting in sharper writing. It also frees up writers to spend more time on the human element of the piece, crafting personal anecdotes or adding humor.

This method does take some trial and error. One programmer who blogs under the pseudonym Gwern Branwen wrote, "'Prompt programming' is less like regular programming than it is like coaching a superintelligent cat into learning a new trick: you can ask it, and it will do the trick perfectly sometimes, which makes it all the more frustrating when it rolls over to lick its butt instead — you know the problem is not that it can’t but that it won’t.”

Whatever program you use, try to get familiar with its quirks. It takes clean and clear writing to prompt the program to give you something you can work with.

Embracing the Future

Marketing is ever-evolving, especially the relatively new field of content marketing. AI intelligent enough to write cohesive articles is an exciting development that can reduce the labor involved in content creation.

Copywriting generators have the potential to increase the glut of poor content on the internet and cheapen brands. However, if used with the oversight of a skilled writer, content-generating programs have the potential to speed up the process of writing exponentially and allow brands to put humans where it counts.

If your company needs forward-thinking content writers and editors that can work with technology to create the best possible content, contact People First Content today!

Illustration of a mailbox