Technology continues to advance exponentially. We now have self-driving cars, full computers in our pockets, and even artificial intelligence capable of writing novels.
Don’t worry, though. As writers, we’re not out of a job … yet.
Right now, AI is capable of writing in the style of Harry Potter and 50 Shades of Grey. A team in Japan passed a literary contest’s first round with a novel co-authored by AI.
There are over 800,000 books on Amazon right now written by AI, and it has even started to produce autonomous niche news stories for companies like Reuters, USA Today, and the Washington Post.
Though AI is able to write its own words, it still has a lot further to go before it makes any bestseller lists.
Writing production has come a long way already, from the 1600s when content had to be copied by hand, to today, where AI can produce content autonomously. It is an exciting time to be alive, but it can be a bit nerve-wracking when you start to think about what will happen next.
Terminator. Ex Machina. I, Robot. There are more than a handful of ideas about what AI will look like in the coming decades.
Some are more pessimistic than others:
“If [AI’s function] is just something like getting rid of email spam, [then] it [might] determine the best way of getting rid of spam is getting rid of humans.”
— Elon Musk
A strong viewpoint from a leading innovator, but also a worst-case scenario.
The machines are getting smarter every day, however.
Does that mean we should stop writing? Should we sharpen our quills and prepare for war? Put down our pens, remove our hands from the keyboard, and start new careers?
NO! AI hasn’t won. There is a lack of humanness to it. It doesn’t understand what it is to be human, and therefore lacks a certain human touch.
Plus, doesn’t writing fulfill something deeper inside you? Even if AI did start to write better than you, isn’t there already someone out there who writes better than you, and yet you choose to keep writing anyway?
What if you had some help along the way? A little AI friend who could pound out some content for you and do the grunt work. That might be okay, right?
AI has made considerable progress in the past decade, to the point where it can now be used as a tool in the writing process. We will see integration instead of segregation. As writers, we still have a place at the writing table. The innovations made are working for us, working on the menial and tedious tasks like data collection so we can do our jobs better, be more accurate, and get more done.
AI has entered the fiction arena. It’s not quite a humanoid AI typing away at a computer. Rather, the work all happens behind the scenes, with the AI program able to generate words in the same way it computes the first million digits of Pi.
That is to say, it needs some sort of input to work with first.
Say you take the first chapter of Harry Potter and the Goblet of Fire and you give it to an AI program to analyze. After the analysis, the AI program will be able to produce sentences that resemble those found in the chapter.
It’s a fun little experiment if you’re interested in seeing snippets of an author’s writing. It also gives you a deeper look at the mechanics behind style.
Longer samples—such as the entire Goblet of Fire book—will produce higher quality sentences and chapters, and can even lead to AI authoring an entirely new book in the sample style. This is what a frequent contributor to Medium and technology enthusiast Max Deutsch calls “Deep Writing.” He describes the process succinctly on Medium:
“1. You show a computer some sample text (for example, the Harry Potter books).
2. The computer identifies all the unique words in the sample text.
3. The computer groups words based on how often they appear together in the sample text (using a particular mathematical model). This is the “learning” portion of “Deep Learning”.
4. You pick a starting word (for example, “The”).
5. Using what it learned in Step 3, you ask the computer to guess the word most likely to come after the starting word (i.e. “The”). This is recorded as the second word.
6. Then, based on the first two words, you ask the computer to guess the third word. And so on.
7. Eventually, you tell the computer to stop guessing after many words, and you have successfully created your Deep Writing.”
— Max Deutsch
The program Deutsch created to do this “Deep Writing” uses a combination of code from Sung Kim, a computer science teacher at the Hong Kong University of Science and Technology, and Google’s machine learning library, TensorFlow. Using the above Deep Writing logic, the program will imitate writing based on text you provide, given it has enough text to work with.
Shakespeare lives again. J. K. Rowling can write 1000 more Harry Potter books. We can finally find out how Game of Thrones ends! But no, that’s where this logic falls apart.
Maybe Romeo can reunite with Juliet over and over. Maybe Harry Potter can repeat grades 1 through 7 ad infinitum. But the AI program won’t be able to create new story lines.
Deutsch’s Deep Writing AI can analyze and produce imitations, but it is no good at creating for itself. It needs raw data to work with and replicate. George R. R. Robot does not exist without George R. R. Martin.
Another example of this—and one this writer finds highly amusing—is a text generator for 50 Shades of Grey.
Lisa Wray bet her friend that 50 Shades of Grey was so basic, AI could write it. She won the bet and dedicated a website to it. If you’ve ever wanted an endless supply of 50 Shades, your wish just came true.
Here’s one of the less raunchy samples:
“I can see nothing, all I can hear is my rapid breathing, making me writhe. His hands gently run down my neck. I groan, frustrated. How can he turn me on with one word?”
It’s not perfect, but impressive nonetheless. An AI program wrote that, and much, much more.
Absorbing E. L. James’ style, the 50 Shades text generator is able to produce fresh imitations on demand, just like the AI Deep Writing process described above.
Input is entered, analyzed, and spat back out in likely ways. It is similar to the suggestions your phone provides when you text. There are words and sentence structures you use often, and the program guesses what word you might type next based on what it has seen you use before.
Give the AI program a framework, along with some of your common sentences and wording, and you may just get an AI program to write a novel for you.
In Japan, a novel co-authored by an AI program titled The Day A Computer Writes A Novel, passed the first round for a national literary award in 2016. The Hoshi Shinichi Literary Award is open to AI-generated novels, though 2016 was the first year they received any AI submissions. Unbeknownst to the judges, AI wrote 11 of the 1450 novels submitted.
The novel is about an AI program that realizes its own budding writing skills and decides to separate from its human counterpart. A bit on the nose for sure, but a fitting subject for an AI-written novel entering a writing competition.
Hitoshi Matsubara, a professor at Future University Hakodate, led a team that authored the human component of the co-authored novel. They provided guidance on plot, characters, and prepared sentences. The team provided about 80 percent of the framework and the AI program then wrote the novel.
They did not win the prize, but passing the first round and beating actual humans is nothing to overlook. It is a significant achievement for our digital counterparts, and a suggestion that perhaps one day an AI program may win first place.
For now, the AI writer is a teammate. There’s plenty that doesn’t come out quite right, as you can see if you look at the program by Deutsch or Wray’s 50 Shades text generator.
As a writer, you could use these tools to evaluate your own writing.
Say you put your own words into a program like Deep Writing and tell the AI program to write a paragraph or a chapter. With a sufficient amount of guidance and a clear framework, you could potentially use it to create your own novel, as the team did in Japan.
You could use the AI writer to see how your style looks from a new perspective. Reading something that an AI program has written in your style could show you what words you use frequently, or maybe overuse. It might reveal that your writing sounds similar to an author whose works you read a lot.
It may even be profitable to let the AI program generate the text and then edit it yourself. Say, for example, you let it write a paragraph or two to start an article or chapter. It might be able to create a sentence that fits better and sounds exactly as if you had written it.
It could spark a creative flame that could lead to a breakthrough in your own writing. This would probably work best if you gave the AI program a large sample of your own work for it to analyze.
From this sample, it would gain a pretty good idea of how you write and could then take examples from across several different stories to produce something that looks new. The setting might change in a subtle but powerful way. Characters might interact or converse in ways you haven’t thought of.
These are just a few of the interesting ways to incorporate technology into your creative life. AI is not only of interest as a tool for the fiction writer, however. It has also been successfully integrated into the business writing world.
AI is good at imitating what already exists. It is good at grunt work and producing large quantities at speed. It is also far better than any human at dealing with enormous quantities of data. As an ally, businesses have started to integrate AI into their team.
AI works for the Washington Post. They have nicknamed it Heliograf, and it covered the Rio Olympics and the U.S. Presidential election.
Heliograf captured results immediately, selected an appropriate angle, and published its news stories before a human writer even had the chance to open Microsoft Word.
Writers at the Washington Post select phrases ahead of time, much like the novel-writing AI program in Japan, and the AI program puts together the appropriate information.
Heliograf’s story on Darrell Issa winning California’s 49th congressional district seat is an example of its prowess. As soon as the race ended and Issa won, Heliograf wrote the article. It chose from pre-selected phrases tied to different outcomes, such as “Republicans retained control of the House,” and “Democrats regained control of the House,” and then combined these with voting data collected from VoteSmart.org.
The article was in the Washington Post voice and was first on the scene.
“Republicans retained control of the House and lost only a handful of seats from their commanding majority, a stunning reversal of fortune after many GOP leaders feared double-digit losses.”
Easy to read, accurate, and efficient. The goal for the Washington Post is not to replace their human journalists with AI. In contrast, their use of Heliograf allows the human journalists to focus on the bigger stories that require more in-depth analysis.
The main intention of using AI is to grow their audience.
California’s 49th congressional district is of little concern to the vast majority of us, but it does concern some readers.
With a niche focus and lightspeed writing, articles like these capture a select few at a time. A few readers captured across a wide array of different niches turns into a lot of readers quickly.
In November 2016 alone, covering the U.S. Presidential election, Heliograf wrote 500 articles, pulling in over 500,000 clicks. These are only a small portion of the over 1 billion clicks the Washington Post averaged, but this is just the beginning.
Heliograf also reports anomalies it finds in the data. The AI program sends a message to Slack, alerting journalists of a possible lead. The human journalists can look at the anomalies and discover a trend emerging as soon as it happens.
This saves the journalist and news agency time, as the alert is instant. This advantage could be the difference between the Washington Post breaking a story and one of their competitors beating them to it.
The Washington Post also uses their AI to keep figures updated in their articles. Stories can stay up-to-date with the most recent information if there are new developments, or if the facts change.
Other news agencies are taking notice.
Reuters uses a tool called “News Tracer” that investigates the integrity of a tweet by looking at confirmations and denials throughout the twitter-verse. They found that about 20 percent of news breaks on Twitter first, but sorting between what is fake and real can be confusing.
News Tracer sorts through the over 500 million tweets that are sent every day to find the new stories, verify their integrity with an in-house algorithm, and alert journalists of breaking news. It’s a great way to get a scoop before your competitors.
USA Today uses AI called “Wibbitz” to create short videos based on news articles and video footage with a synthesized voice-over.
These do not sound like robot overlords. These are new employees who can do the tedious tasks with unheard-of efficiency, accuracy, and ability, that don’t need a steady stream of coffee and a salary to get the job done. (But you may have to reimburse the developer. He needs coffee.)
Other content businesses are utilizing the same AI techniques. A company called ICON Group International, Inc. has 800,000 books up on Amazon right now, all written by AI.
The man behind the madness is Philip M. Parker. A professor of marketing at INSEAD Business School, he has been in the AI game for 10 years already, filing a patent for his software in 2007.
The AI program writes books on demand for a specific niche. Subjects include foreign language dictionaries, business reports, crosswords puzzles — anything that is data driven.
The AI program takes about 20 minutes to source the proper material and produce an appropriate book. I can’t think of any authors who can write a worthwhile article in 20 minutes, let alone an entire book.
Parker designed the program to look for formulas in the content. It breaks down the data into chunks, lays out the formula, and from deconstruction, book-writing begins.
Each book costs between $0.20 and $0.50 to make and sells for anywhere from $20 to $350. I don’t think the AI program is looking to save money for new shoes, meaning Parker and his company ICON Group stand to make substantial profits.
AI is perfect for writing in the business world. As a tool for growth and efficiency, it helps companies create content that is not only intelligible, but which there is demand for in the marketplace. AI creates content fast, has an insatiable data appetite, and allows human writers to work on the more in-depth stories, creating higher profits and less waste for the company.
There can be a lot of grunt work in a news agency or online business. The sheer amount of data available these days can be daunting to sort through. This can put content producers behind when they need to get their stories and information out quickly.
AI can help minimize the grunt time, and help the writer focus on more important details that require a human touch. AI might not know what news is most relevant at the time, so although it can produce an article quickly, it could still miss the mark. While AI is efficient with the data, humans are efficient and accurate with the analysis.
Together, the information can be found, analyzed, and made available quickly, providing an edge for a business. In our fast-paced world, that edge could mean the difference between sinking and swimming.
With so much information available, it can be hard to tell the difference between fake and real information. As with Reuters’ News Tracer, AI can help find the truth.
The last thing a business needs is to be found producing fake information, as it is an advantage to be seen as a reliable source.
By quickly providing accurate information, AI can help a business become a leader in their particular niche. The human and AI work together to get the information to the right eyes and grow their audience, reader by reader. As the audience grows, so does the business.
Collaboration between humans and AI can allow each to focus on their strengths. The human provides the curation, framework, and human point of view, and the AI program provides the data and speed.
Right now, an AI program can imitate a fiction writer’s style, producing content that resembles how the original author might have written it. It can even write its own novel, given a team to guide it along the right path.
AI helps news agencies like the Washington Post and Reuters attract readers in niche areas, thus growing their audiences. It also allows them to retain their human journalists for more in-depth work. This saves both time and money, while providing growth for the business.
Technology seems to be on an inevitable track towards true Artificial Intelligence. By that, I mean an AI program that thinks like a human and therefore can write like a human.
I personally don’t believe AI will eliminate humans in order to eliminate spam, like Elon Musk thinks it will, just as I don’t think content-writing AI will think it’s a great idea to eliminate the internet in order to eliminate competition.
The thing about AI is that it will always be artificial. Will it ever feel the elation of a good writing day, or the self-loathing that comes when your hands are frozen on the keyboard?
Probably not, but then the question becomes whether those factors are important in the writing process. Are those emotions necessary for art or content that goes deeper than the data?
In the coming decades, we will learn the answers to these questions.
For now, AI can help writers get their work done faster and allow them to focus on the more important aspects of the craft. With each new iteration of technology, this trend of integration seems to be on the rise.
It is a trend of empowerment. The writer and AI can work together to get more done, grow their audiences, and do better work.
Garrett Grams is a freelance and fiction writer. He loves the SciFi and Fantasy realms. After graduating with an IT degree he joined the cubicle masses. His soul cried out to pursue a life in writing, and so at the end of 3 years, he left for Vietnam where he taught English for 18 months. He is currently working on a dystopian novel that WILL be finished this year. He can be found in the back corners of coffee shops in a crazed caffeine-addled typing fit, or on the interwebs at his website.