Note to readers: This column's algorithmic writing program is experiencing temporary technical difficulties. Please excuse any grammatical errors or illogic that may result.
Well, that's my story, and I'm sticking to it. And perhaps we should start getting used to the idea of computer-generated writing, because it's with us, and not just in theory. In fact, it's becoming increasingly hard to ignore — in no small part because human writers and editors are so eager to spin real-life examples into attention-grabbing, usually ironic headlines. (Note exhibit above. Guilty as charged.)
Take two instances that occurred in March. There was much ballyhoo about a study by Swedish academics that found that subjects who were asked to read two NFL game recaps — one produced by "generative software," the other by a newspaper reporter — were "not able to discern automated content from content written by humans". Almost concurrently, another glut of coverage ensued after Slate wrote about The Los Angeles Times running computer-generated first reports on earthquakes (made possible by a "Quakebot" algorithm written by a reporter / programmer at the newspaper).
In both cases, writers whose functions could, over time, quite conceivably be whittled away or usurped by computers laid on the gallows humor. For the Quakebot story, the ever-restrained Huffington Post declared: "It's All Over: Robots Are Now Writing News Stories, and Doing a Good Job."
Wired UK's piece on the Swedish study bore a nearly identical headline ("Robots Have Mastered Writing. Goodbye Journalism"), although the article itself was solid and nuanced, duly reporting the study's decided limitations. That was not true of many other media outlets, pointed out Columbia Journalism Review (CJR), which chastised journalists for ignoring the study's serious methodological flaws because the story was "too good to check and the clicks were too easy to garner".
Another underlying irony is that computer-generated news coverage is actually old news. In April 2012, Wired itself reported on Narrative Science, a Chicago-based company that "trains computers to write news stories". (The company's site explains that its patented Quill technology mines clients' data and, "using complex artificial intelligence algorithms, extracts and organises key facts and insights and transforms them into stories, at scale.")
Wired's Steven Levy wrote that Narrative Science-generated stories covering everything from sports to corporate financial results to the US presidential race were already being run on the websites of "respected publishers like Forbes, as well as other internet media powers (many of which are keeping their identities private)." He continued: "Niche news services hire NS to write updates for their subscribers, be they sports fans, small-cap investors, or fast-food franchise owners. And the articles don't read like robots wrote them."
Robonews
Most chilling, at least for editorial types, were quotes from Narrative Science's CTO / co-founder, Kristian Hammond. On one hand, he insisted that robonews won't eliminate human writers, but instead dramatically expand the universe of newswriting, as (in Levy's words) "computers mine vast troves of data to produce ultracheap, totally readable accounts of events, trends, and developments that no journalist is currently covering." On the other hand, when pressed, Hammond prognosticated that more than 90% of all news will be computer-written within fifteen years.
Of course, that may be wishful thinking on his part. But I would submit that we're already laying the foundation, by creating human robo-writers. Since online engagement metrics became the coin of the realm (and the key to retaining a modicum of value to one's employer), humans have become adept at generating headlines and copy engineered to maximise traffic, clicks and social sharing. Indeed, we have "scientific guides" for honing these "skills".
Unfortunately, this is likely to prove a short-lived survival strategy, because even humans who know all the tricks don't always produce the most effective clickbait copy or meme of the moment. That's why, despite well-known "best practices" - heads and tweets with proven, formulaic structures and endlessly recycled hot-button words / phrases (eg. "accuses", "senseless violence", "faith in humanity") and cliffhanger promises ("You Won't Believe What Happens Next", "… It's Not What You Think") - the big clickbait sites routinely test scores of headlines to find the optimally titillating or annoying combination. Resistance is futile.
The logical next step is bypassing all that by employing algorithms engorged with continuous, real-time data on what's actually generating online activity among one's target audience or the public in general —programs that can respond and spit out copy in minutes, without missing deadlines, or complaining about long hours, or expecting benefits.
No, I'm not likening the websites of (most) mainstream magazines and newspapers to BuzzFeed and its ilk — or in any way minimising the value of the information, insights and excellent writing that so many deliver in prodigious quantities, day in and day out.
But the battle for advertising / marketing dollars continues to escalate, particularly now that a large and rapidly growing chunk of marketers' spending decisions are going "programmatic". (Direct ad salespeople fear automation, too.)
Strategic responses to these pressures include focusing investment on driving audience metrics most expediently and cost effectively, while rooting out ostensibly impractical or obstructionist old-school values and practices with hard-to-quantify ROI; offering ever more alluring "native" content opportunities for marketers; and leveraging fiscally attractive content alternatives (like replacing some staff with freelancers / contributor networks). Automating whatever can be automated without undermining the value of media brands would seem an inevitable strategic component going forward.
The key phrase, of course, is "without undermining brands". From a strictly bottom-line standpoint, the outcomes of evolving strategies will hinge on corporate leaders' ability to judge which investments and directions will result in long - as well as short-term fiscal health, and where to draw the line.
Which media companies will prevail? My bet is on those that elect to invest significant portions of the cost savings, and revenue, that they realise through expanding technological capabilities, smaller headcounts and new brand extensions, in nurturing and rewarding exceptional writers and editors. Meaning talents who can deliver the audiences that marketers want by producing exceptionally relevant, compelling (and yes, entertaining) content that sets their brands apart from the undifferentiated, lowest-common-denominator clickbait Borg.
