Content! Consumers demand it, but it takes so long and costs so much. But what if you had an employee who could create 2,000 articles per second? One who didn’t make typos? Or who didn’t even need bathroom breaks – or a salary?
Content created by an artificial intelligence (AI or neural networks to be specific) may seem like a dream come true to every manager who’s had writers miss deadlines (and even the mark) entirely. Make a few keystrokes, enter some key phrases, and voila: A completed post or page hot in an overworked marketer’s hands.
But technology on its own isn’t the answer to society’s problems. While artificial intelligence is here to stay, the chances that it will replace all of our jobs seem to be pretty slim.
AI is Already Producing Thousands of Articles
Though many of us think of AI as the future, firms like data science and software company Automated Insights are already leading the AI-generated content charge. The Associated Press has used Automated Insights’ Wordsmith platform to write 3,000 quarterly earnings reports each quarter.
Wordsmith has also produced millions of articles for companies like Allstate, Comcast, and Yahoo. As you browse financial and fantasy-football sites, watch for a line at the end: “This story was generated by Automated Insights.”
content automation has also proven useful in hunting down the most popular related articles
The kicker? The automated system produces fewer errors than the reporters who used to put these admittedly dull financial or statistics-based pieces together.
Beyond the writing, content automation has also proven useful in hunting down the most popular related articles. Machine learning and automation promises to find these pieces and then use algorithms to analyze engagement, platform, time posted and more.
It can analyze the most popular articles and replicate their style, sentence length, and topic to produce a similar, winning piece of content.
The suggestions Wordsmith delivers to stakeholders promise to save considerable amounts of money and labor. In the Content Marketing Institute’s 2017 B2B Content Marketing report, 52% of content marketers surveyed listed a lack of time as the top hindrance to their content marketing success.
Could neural networks and artificial intelligence be the answer to the time crunch that many content marketers face?
What’s a Poor Content Marketer to Do?
You may have heard about how robots, automation, and software will make many occupations obsolete over the next 20 years. Truck drivers may be replaced by autonomous vehicles, financial and insurance analysts by software, and bank tellers by ATMs. Will content marketers be next on the chopping block?
Luckily for those of us who make a living adding to the Great Encyclopedia in the Sky, neural networks have their own Achilles’ heel; they do well with facts but freeze in the face of feelings.
email, social media, and content marketing depend heavily on the human-to-human connection
Most copywriters learned the “capture them with emotions; convince them with facts” model not only of persuasive sales writing, but all writing designed to engage readers. Persuasive marketing language presented in the proper medium springboards from feelings like desire, excitement, sorrow, fear, envy, greed and more. It also leverages word play, a nimble tactic that seems as yet outside of a neural network’s digital grasp.
As more content is produced by machines, humans won’t bother opening their emails or reading those posts. Email, social media, and content marketing depend heavily on the human-to-human connection. People want to interact with other people on their digital channels, not read canned phrases strung together via an algorithm. People go to their email and social media feeds during the day to feel human.
Facebook’s status updates and cat videos serve as a refreshing tonic of feelings amidst a rote, computer-dominated day. Companies who automate blog posts, articles, web pages and even email newsletter content do so at their peril. We’ve found our best-performing content is powered by passion, a distinct voice, and a supremely unique point-of-view.
AI Can’t Fully Replace Marketers…
From time to time, humans test whether artificial intelligence can be as witty and deep as their content marketing colleagues. The results may surprise you.
Research scientist Janelle Shane entered 2,000 proverbs into a data set and asked a neural network to write its own. (As a refresher, a proverb is an oft-repeated bit of conventional wisdom like “fortune favors the bold,” or “the squeaky wheel gets the grease.”)
These sayings make perfect sense to us, but trying to code the personification and other sly literary devices that make proverbs cool gets tricky for a system operating on ones and zeros. After all, a human writer relies on a lifetime’s worth of social interactions and experiences that help guide word choices, warm up to readers, and avoid faux pas. Still, neural networks have progressed far enough to drive cars, write music, and defeat champion chess players. Is writing that different from those other endeavors?
So what new proverbs did the computer produce? Here are a few:
- “A mouse is a good bound to receive.”
- “Do not come to the cow.”
- “A good face is a letter like a dog.”
While these phrases do indeed have the rhythm of proverbs, they don’t offer any insight. What is the life lesson to take away from “Do not come to the cow?” Should the cow come to you?
Shane also programmed a neural network to create new paint colors and append a name to each. Granted, real human-named paint colors like “Dragon’s Blood” and “Mayonnaise” may not exactly appeal, but the AI went even farther into the ditch. Shane fed it 7,700 Sherwin-Williams paint colors, matched to their RGB values. Feast your eyes on these tones:
- “Sane Green” for a purplish blue.
- “Dondarf” for violet.
- “Gray Pubic” for sky blue.
- “Bank Butt” for a grayish pink.
- “Stargoon” for a mustard tan.
Shane agreed that the network completely missed the point. No one’s room will be painted in “Bank Butt” no matter how much they love dusty pink.
…But AI Content Does Have a Place to Succeed
The bottom line is that the millions of articles that programs like Wordsmith create are very dependent on formulaic source material. Do you know the formula of a quarterly report?
- description of Program/Initiative
- append Income Statements, Cash Flow, and Balance Sheets
Anyone who’s ever received one knows that the quarterly reports for Quarters 2, 3, and 4 are very similar to Q1.
Each of the elements above goes into a spreadsheet. Patterning after the human-generated template (maybe Q1), the neural network puts each piece in the right place and adds a little fluff to knit the pieces together. It also leans heavily on the thesaurus for some good synonyms as it goes from quarter to quarter for each company.
Similarly, the articles Wordsmith produces for Yahoo and Comcast are both formula-based and figure-related. The introductory sentence, transition sentences, and conclusions can be coded into a neural network, again depending on the wording used to sound fresh in each iteration.
Certainly AI-generated content has a place. It may even be providing a great service to content writers and even journalists bytaking over the role of compiling the often-dreaded first draft.
the use of Wordsmith has not resulted in writer layoffs
Jeremy Gilbert, director of strategic initiatives for The Washington Post, recently told Wired that AI gets the basic facts down which frees journalists to create content with deeper analysis and greater meaning. “If we took someone like Dan Balz, who’s been covering politics for the Post for more than 30 years,” Gilbert said, “and had him write a story that a template could write, that’s a crime.”
Any good journalist or content writer can appreciate an AI setting down the facts in a logical order, freeing the reporter up to add flourishes, draw conclusions, and make connections to other articles and even societal trends.
Even those who oversee the AI-generated quarterly reports at the Associated Press aren’t quite ready to dispense with living, breathing input on those dull documents. In fact, the use of Wordsmith has not resulted in writer layoffs.
Philana Patterson, assistant business editor at the AP, tells The Verge that, far from taking writers’ jobs, computers free writers to add more robust insight and a connection to the bigger picture.
“One of the things we really wanted reporters to be able to do,” she explains, “was when earnings came out to not have to focus on the initial numbers. That’s the goal, to write smarter pieces and more interesting stories.”
Most writers jump at the chance to include more information, but time and budget limits prevent them from making the grand statements that make the piece so much better. AI programs could help with that.
Content Marketers: Don’t Fear AI – Embrace It as a Helping Hand
We have the sci-fi genre to thank for making AI sound scary, and the media, driven by the “if it bleeds, it leads” mindset, certainly doesn’t help. Marketing news is just as guilty of using fear to grab eyeballs, but don’t let the headlines about AI taking content jobs worry you. Just as the invention of the calculator didn’t erase the need for mathematicians and the creation of Photoshop didn’t eliminate the need for photographers, AI writing platforms don’t wipe out the need for humans to provide meaning, creativity, and direction in content creation.
AI-generated content will likely make marketers’ jobs easier and more meaningful; any marketing manager who’s winced at assigning 25 pages of product- or location-based content to a writer knows this. Repetitive pages have the power to scramble even the most experienced writer’s brain. Why waste that energy on something a smart robot can do?
Thanks to that robot’s ability to gather data and hand it to you in a nice (albeit bland) piece of writing, you are now free to harness your creativity, skill, and uniquely human experience to create content that holds more value, explores deeper meaning, and evokes real emotion in readers.
That’s something only a human writer can do.