Can this machine write an article without you?
Automatic Writing – The New Face of Journalism
There is a specter haunting the world of writers – the specter of automatic writing. That’s right, an article or report generated by a computer algorithm without human input. An algorithm is a “step-by-step procedure for solving a problem or accomplishing some end especially by a computer.” This is a natural outgrowth of the science of Artificial Intelligence. So new is this topic of automatic writing that there are very few Google or Bing searches on the subject. People are only gradually becoming aware of it. This will change soon, because automatic writing will revolutionize journalism, not to mention freelance writing. And it will happen quickly.
How’s this for a snappy piece of sports writing?
“WISCONSIN appears to be in the driver’s seat en route to a win, as it leads 51-10 after the third quarter. Wisconsin added to its lead when Russell Wilson found Jacob Pedersen for an eight-yard touchdown to make the score 44-3 .” As with many sports columns, the article, asdiscussed in The New York Times, was written 60 seconds after the end of the game. So what? Why am I telling you this? Well, the article was written by a computer. Yes, computer generated writing is here, and it’s not going away.
As a writer I think I hate this development. Writers, especially freelance writers, are a worrisome crowd. They have to find the next assignment, create a story line, meet a deadline, and oh yes, pay the bills. I hate to give writers something else to worry about, but facing reality is always a healthy thing. Yes, a lot of the staple subjects of article writing will soon migrate to computer algorithms. Full disclosure – I wrote this article myself. I used a computer, but the computer didn’t write it. I will also write any future updates to this article. Maybe.
Article spinning is a type of automatic writing using rudimentary artificial intelligence algorithms. Article spinning means taking an article, putting it through spinning software, and voila – out comes an article that has changed the words with synonyms so as not to engage the wrath of the Google prohibition against searchable duplicate content. Unless the article involves a simple subject in simple language, the outcome can look like yesterday’s corn beef hash.
The Beginnings of Automatic Writing
Automated systems have changed the way we do things since the invention of the printing press. and now, the act of writing itself is becoming automated. Narrative Science, a start-up company in Chicago, is using artificial intelligence to generate articles. The company formerly launched in 2010. It began in Evanston, Illinois as a joint research project with the Northwestern University Schools of Journalism and Engineering. The three founders were Stuart Frankel, CEO, formerly with Doubleclick; Kris Hammond. Chief Technology Officer and professor of Computer Science and Journalism at Northwestern University and the founder of the University of Chicago’s Artificial Intelligence Laboratory; Larry Birnbaum, Chief Scientific Advisor and professor of Computer Science and Journalism at Northwestern. For an interesting description of this development, see: ”Steve Lohr, “In Case You Wondered, a Real Human Wrote This Column,” NY Times, September 10, 2011.
This new company is rewriting the history of journalism. The basic idea of computer generated writing is simple. First, develop a huge database of information on a subject using data mining techniques. Sports and finance are natural areas of inquiry because any discussion of either subject requires a lot of numbers, people, comparisons and history. Once the database is built, then write an algorithm to go in and extract data and put it into intelligible narratives. Using baseball as an example, the algorithm is taught to understand that the most runs scored wins, that an inning is over after three outs and all of the other rules that run the game. Set loose on the huge database, the algorithm soon is able to figure out that batter X has only a 10 percent chance of getting a hit off of pitcher Y, based on the historical information in the database. The algorithm also learns the lingo of the game, so that when it generates a report it says things like Jones “smashed one” over the left field wall, or that a batter was “thrown out looking.” And what’s most shocking is that the algorithm and its pal the database can generate a story within seconds of the end of a game, with journalistic grace and incredible accuracy.
The world of finance, awash in facts and figures, is also fertile ground for automatic writing. A report, written for a business magazine, may read: “XYZ Corp.’s last quarter was a bitter disappointment, with revenues off of its previously stellar chart climbing stats, and profits are also in the tank. Investors will head for the exit door.”
How Automatic Writing Will Change Journalism
Technical writer Steven Levy, writing for Wired Magazine, has penned an excellent article on automatic writing and the future of journalism entitled: “Can an Algorithm Write a Better News Story Than a Human Reporter?” He discusses how the programmers are learning how to make the algorithm figure out things faster. Writing restaurant reviews, for example, requires that the algorithm look at the database of restaurant information and zero in on certain critical metrics like high review scores, good service, good food and a couple of customer reviews. Within hours, according to Levy, the database could crank out pithy little articles like “The Best Italian Restaurants in Atlanta” or “Great Sushi in Milwaukee.” Does this remind you of a HubPages article or a Textbroker assignment? Levy talks about a competitor of Narrative Science that began as a company known as Statsheet, which concentrated on reporting sports contests. As the excitement unfolded, the company founder changed its name to Automated Insights. Levy quotes Robbie Allen, the founder, about its previous thinking that the company would limit its mission to data-rich industries: “Now I think ultimately the sky is the limit.” When interviewing Kris Hammond, the Chief Technology Officer of Narrative Science, Levy asked him what percentage of news articles would be written by computers in 15 years. Hammonds answer may send shivers into the spines of writers. Hammond said “More than 90 percent.” Are the reports accurate? Levy talked to Lewis Dvorkin, Forbes Media chief products officer , and asked about the accuracy of the computer generated articles from Narrative Science. Although reporters are known to get things wrong, he did not find one instance of an error in any Narrative Science articles. Algorithms don’t miss things. Jeopardy fans the world over looked with nervousness as an IBM computer named Watson (after IBM’s founder) took on two previous Jeopardy champions in February 2011. Watson won hands down and sent the two champs packing. Artificial intelligence had hit prime time.
Breakthroughs have a way of expanding and changing things. In the early 1980’s the primordial jungle of the computer revolution, we were amazed at how you could highlight a paragraph or word then copy or cut and paste it. The early PC enabled us to do more with what we had. Artificial Intelligence, on the other hand, goes beyond what we have. For looking up data, correlating it and making relevant conclusions, we can’t compete with algorithms.
Should Writers Be Worried?
Ayn Rand once famously said: “You can avoid reality, but you can’t avoid the consequences of avoiding reality.” Some who may be reading this article may think that the province of writers is safe, that a computer program, no matter how sophisticated its algorithm, can never replace the analysis that a human being can bring to bear on an issue. What about the article that you are reading right now? I have looked at the reports of the amazing new companies on the scene, I have selected a few telling quotes, and I have given it my analysis, which is what I am doing right now in this paragraph. But suppose the Narrative Science people put all of the data they have and that which they can get their hands on into a database devoted to the subject of artificial intelligence as it applies to automatic writing. Do you think their algorithm would not alight upon the quotes with the predictions of automatic writing being responsible for as much of 90 percent of articles in a few years? Do you think that the algorithm can’t look at the numbers and make mathematical projections far better than me? Yes, I do think writers have something to worry about, unless they write strictly for pleasure. Are only nonfiction article writers at risk? A computer program can be stuffed with basic plots and characters, and the algorithm can pick and choose, just like a writer does, and come up with a novel. I have favorite novelists who I read not because I like them, but because I like their writing. Show me an algorithm that can weave a good plot with exciting characters who say compelling and funny things, and I will send him (it?) fan mail.
Will a computer ever win a Pulitzer Prize? Narrative Science’s Kris Hammond thinks so. He referred to a pundit’s prediction that a computer will win the Pulitzer in 20 years, and disagreed. Hammond thinks a computer program will win the Pulitzer Prize in five years (that would be 2016).
What’s a writer to make of this? Do you think a computer could write a sentence like an Ernest Hemmingway, a George Will, a Tom Wolfe or a Joan Didion? I, for one, can’t conceive of this. But then I thought the idea of an online auction site (Ebay) was dumb, and that nothing could replace the keyboard and mouse. So I’m not making any predictions. I’m just contemplating how one sends a note of congratulations to a computer that just won the Pulitzer.
Will computers and their algorithms ever form principled opinions and share them with us? Just ask Hal, the spacecraft computer in Kubrick’s movie 2001(in 1969): “I’m concerned about the mission Dave.”
Copyright © 2013 by Russell F. Moran