Natural Language Generation Turns Robots Into Writers

Natural Language Generation Turns Robots Into Writers

What would our computers tell us if we gave them a voice? We’ll soon find out thanks to Natural Language Generation which gives computers a written opinion on virtually anything.

For now, we must program their responses, but soon they’ll form their own opinions and develop a creative voice. This may seem a long way off, so let’s consider their progression as a writer in comparison to a human.

A child progresses as a writer by starting with basic creative writing exercises: What did you do over summer break? Over time, assignments focus on creating technical, justifiable arguments. 

The progression is flipped for computers.

“I’ll need that business report by 4”

Natural Language Generation already has the basics down. With every word in the human language at its disposal, computers only focus is arranging these words. 

So, why not start with an arrangement of words that is so formulaic, writing them is literally just plug and play.

I’m referring to business content. Whether it is earnings reports, SWOT analysis, or market research, with a template and data on hand, Natural Language Generation can spew out a piece of business content that rivals that of a human.

Don’t believe me? Well, a Natural Language Generation tool called Quill writes Forbes’ annual earnings reports. In fact, Gartner predicts machines will author 20% of all business content by 2018.

And that isn’t the only at-risk content.

Anderson “The Natural Language Generation” Cooper

Flashback about eight months to the Rio Olympics. With hundreds of events, exciting matches happening simultaneously, and incentive to be the first to report the latest medal count, news outlets turned to robot journalists for speedy coverage.

Over the 15-day event, Chinese media outlet, Toutiao, unleashed its secret writing weapon: the Xiaomingbot. With its digitized pen in hand, the Xiaomingbot wrote over 450 articles…that’s 30-40 a day!

They weren’t alone either. The Washington Post employed Heliograf, their own robot writer, which they repurposed for coverage of the 2016 Presidential Election.

While neither Heliograf nor Xiaomingbot are capable of writing introspective journalistic pieces, they gave us the news, free from opinion. This allowed other journalists to dig into deeper stories. Perhaps foreshadowing a future where Natural Language Generation software covers all news. 

The secret sauce…

Why are business and news content so appealing to robot writers? Actually, because these forms of content lack emotion, they are easy to write for a robot writer.

We expect emotionless news and business content. Think about how catastrophic it would be if Donald Trump wrote The Trump Organization’s earnings report in a tone of resentment after having a bad hair day. Or, if Anderson Cooper started cursing at terrorists for being too extreme.

It’s a natural fit for an emotionless computer to write with no particular emotional tone. For this reason, people believe that computers will never master creative writing, which relies on the outpouring of one’s emotions.

But, what if Natural Language Generation could fake creativity?

The New York Times, Stephen King Edition

When you take a photo with your phone, you have the option to add filters afterward. Maybe an orange-hued filter to make the photo look warmer, or a glitchy filter to make it look futuristic.

Well, Natural Language Generation can add filters to words. One of the first books written solely by a computer was, True Love. Essentially, it took the famous book by Leo Tolstoy, Anna Karenina, and worded it in the style of a Japanese author called Haruki Murakami.

After analyzing hundreds of thousands of words from a single author, a computer effectively understands their writing style and can mimic their writing patterns. Now, apply this technique to other written work. 

Imagine buying the Stephen King filter for New York Times articles – reading the gruesome, grisly tale of today’s events. Or, attaching a Dick Vitale filter to Bloomberg reports – listening to the stock news with that splendid, superb announcement style.

Natural Language Generation can turn boring news into interesting works from our favorite personalities.

Computers are finding their voice

In the process of learning other author’s styles, Natural Language Generation software may develop its own style.

All good writers go through a period of finding their voice. To do this, they read thousands of sentences from works they love, emulate the authors they admire, and eventually develop their unique style.

Many people who’ve covered Natural Language Generation believe that a computer will never think creatively on its own – that we’ll never read a profound piece of literature written by a computer. But that’s a limited mindset. The Day A Computer Writes A Novel, a book written by AI almost won a literary prize in Japan.

As best portrayed in the movie Chappie, a brilliant programmer creates true artificial intelligence that can think on its own, pitches it to the CEO, and gets the following response:

“Do you realize you just came to the CEO of a publicly traded weapons corporation and pitched a robot that can write poems?”

Most of the money poured into AI infringes on your digital privacy. Resources aren’t allocated to developing creative artificial intelligence. However, just because we haven’t seen it yet, doesn’t mean it’ll never happen.

All of the examples I’ve mentioned above are the prelude to the album of creative Natural Language Generation to come.

On 2016’s New York Times Bestseller list was The Hidden Life of Trees by Peter Wohlleben, which attempted to uncover the intelligent networks of communication between trees. I’m sure any trees that read Peter’s book laughed at how little we know of them, just like we laugh at the elementary writing abilities of Natural Language Generation.

In due time, Natural Language Generation software will complete its research phase, discover its creative voice, and amaze us with its first introspective novel.

Fake it ‘til you become it

Everyone knows the old adage, “Fake it ‘til you make it.” Pretend you are “one of them” until they accept you. But, in my sophomore year of college, a professor explained to me how horribly wrong this saying is.

If you fake it until you make it, no change is made, and you are the same person just in a new environment.

Rather, he said, “Fake it ‘til you become it.”

If your goal is to get better at giving presentations, show the signs of confidence until you feel confident in front of an audience. If your goal is to become a better friend, pretend to care about all your friend’s problems until you actually do care to listen.

Goals can only truly be achieved internally with an improvement of the spirit. All else feeds the ego.

This especially holds true for learning. Memorizing and remembering things isn’t understanding. Learning occurs when one internalizes the lesson and relates it to their own experiences.

With Quick Theories – my weekly newsletter on the future of technology – I give you the facts with my opinion but want you to form your own thoughts.

If you enjoyed this article and would like to receive one like it every week, sign-up here: quicktheories.com

  1. Pain is it’s mother. Strengths​, weaknesses, opportunities and threats are it’s food. Analysis is it’s compass. Writing is it’s purpose. For these reasons, creativity can only be modeled, never developed, by artificial intelligence.

    1. But, is any work of art truly original? Isn’t creativity just a culmination of other people’s influence on the artist?

      Humans develop creativity by looking at those who came before them, learning their way of expression, and copying styles. Experience is secondary. In other words, don’t we all model creativity?

      Writing is nothing more than placing words in a particular order to make others think. If a computer can create a contronym phrase that makes me ponder my existence and change my mindset, then that’s creativity in my book.

      1. Dear Dr. Semantics,

        If your hypothesis is true, why is Lyman Frank Baum’s reputation as a creative writer unnoticed until someone tells you he wrote The Wizard of Oz?

        While there are millions of girls with guitars, even people who hate country music will tell you how Taylor Swift got started.

        If your innocent cell mate, once sentenced to 3 consecutive life terms in prison, was then exonerated through DNA evidence, would you quietly remain an innocent prisoner, or would you fight for the newly-first-time-discovered creativity of scientific evidence?

        The first person quoted saying “there is nothing new under the sun”, became an historical icon of depression.

        Semantics will get us no where when debating in someone else’s blog post.

        When a computer creates a contronym phrase that changes your mindset, I will invite you, and Qu, and the computer over for a couple beers, which I promise will fry the computer’s motherboard, while helping us get to the bottom of semantics.

  2. Wouldn’t it be great if robots took over news sites? That ways, hundreds of people dying in Colombia floods would be given more news coverage than a few people dying in Stockholm.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Up Next:

Artificial General Intelligence Approaches Singularity

Artificial General Intelligence Approaches Singularity