John Henry was a steel drivin’ man, digging tunnels in the mountains in West Virginia. He was the best there was. He would rear back with his hammer until it touched his heels, and drive steel spikes into the rock with one mighty blow, so the holes could be filled with dynamite, and the tunnels could be dug out. No one could work as fast or as well as John Henry.
One day, the big bad bossman told John Henry that he was going to be replaced by a steam driver, a monster machine that could outwork any man. John Henry told the boss that no machine could beat him, and he would die with a hammer in his hand.
A race was set up between the two, man versus machine. The steam driver drilling holes into the rock, and John Henry hammer-slamming spikes home. The two combatants went at it so hard and so long, no one knew who was going to win that day.
Once the whistle blew and the time ended, not only had John Henry won the battle, the machine overheated and exploded from all the effort. A few seconds later, John Henry’s heart gave out, and he died, still gripping his mighty hammer.
While John Henry proved he was the better man, progress never stops. The machine was fixed, and the workers were still replaced in the end.
The Machines Are Still Coming
A couple years ago, I wrote about how the content shock will flood the Internet with cheap, poorly-written content, which will make it harder for good content marketers to get their stuff actually read.
Except now it’s worse.
While writers have only had to worry about competition from other humans, now it’s computers that are able to write. And if anyone is going to be creating the content shock, it’s going to be machines that can turn out articles in seconds and minutes, not hours.
This is what worries me. Not the poor writers who blob together a few sentences that would barely pass high school English. But the machines that can actually do an acceptable job of it.
I worry about companies like Automated Insights or Narrative Science, creators of software programs that can automate writing. Automated stories like Narrative Science’s stories for Forbes about earnings previews of publicly traded companies. Or Automated Insight’s mechanically generated stories for fantasy sports leagues. All output is based on algorithms and formulas, and is built on the principles that made Mail Merge so cool in the 1990s.
Just dump in the data, hit the button, and the algorithms will select language from a vast dictionary of phrases based on differences in scores. Once you’re done, you have a fact-based article about how one team fared against the other, how this quarter’s results are better than last quarter’s, or what your web analytics actually mean month over month.
Now, a new piece of software, Articoolo, can take a few keywords, scour the Internet for other articles about that topic, and create one that gives you the gist of what else is being said on the Internet. It’s not great writing, but it’s “good enough.”
Not only is it mechanical and soulless, it still falls into that “mediocre” category that people have come to accept. They accept it because we’ve been conditioned to by all the crappy writing that’s come before it from people who don’t give a shit about the quality of their work.
The Machines Are Improving
The content created by computers now is much better than what was being plopped out just a few years ago. And it’s getting better, which is worse for human writers. In 2012, Kris Hammond, CTO and co-founder of Narrative Science, told The Atlantic that it’s “theoretically possible for the platform to author short stories,” although The Atlantic author believes it will never match the soul and emotion of a human-generated story.
(Does it matter? By all accounts, 50 Shades of Grey was poorly-written, but still earned nearly $100 million in 2013. So much for the soul and emotion of human writing.)
Combining a slackening of acceptable standards with an improvement in robot writing, and this is where most of the content flood will come from in the next five years. Hammond once told Wired magazine that 90% of the content on the Internet will be generated by automated writers by 2027.
While I worry that it means fewer humans will be writing content, Hammond says that’s not the case. Instead, it will be because the machines are generating more and more articles than ever before.
And that’s where Schaefer’s content shock is going to come from.
The John Henrys of the written word are facing the new-fangled steam drivers, and it’s about to get ugly. As decision makers lower their expectations about what’s good writing, that means they’re more willing to accept bad writing, or not-quite-human writing. It means that people will blindly accept writing that wouldn’t have passed muster 50 years ago, but is considered “good enough for who it’s for.”
I hope it doesn’t mean we’re going to die with a pen in our hands.
Photo credit: Gene1138 (Flickr, Creative Commons)