Subject. Verb. Object.
These are the basic building blocks of written communication. It’s what you need to make a complete sentence like the one you’re reading now.
Structured. Logical. Direct.
This also is why parts of my chosen career are ripe for takeover by robots.
For millions of Americans, the defining realization of how fast artificial intelligence is evolving came in 2011 when Watson — IBM’s now-celebrated language processing computer — won the popular TV quiz show Jeopardy by beating two of the game’s top champions.
I watched with fascination as well. But, for me, the point was driven home even harder a few years later with an email from colleagues at The Associated Press, where I moonlight as a remote sports writer. The world’s largest independent news gathering organization had just announced that the robots it had begun using to produce standard business blurbs such as corporate earnings reports were being modified to write quick recaps of competitive sporting events.
Samples of robot-produced articles were distributed alongside human-written versions for comparison and it was indeed difficult in many cases to figure out which was which.
Although still rudimentary in form, and largely limited to formulaic types of articles that rely on standardized data, it was an undeniable glimpse into the growing sophistication of AI.
Journalism, and writing in particular, had long been considered a career field that was basically shielded from automation because of the creativity required.
But now there’s an app even for that.
And it’s not just business and sports news.
In the November 2016 election, the Washington Post used a robot to produce near instantaneous articles about the outcome of congressional races as soon as results were posted. USA Today is experimenting with robots that automatically convert articles into short online video segments complete with pared down scripts, narration, and images.
All say human writers will remain key to an overall newsroom but envision a future where robots shoulder much of what’s known as the grunt work, basically fast reports on data-heavy topics.
The Associated Press calls it augmented journalism.
In a new report describing how the wire service envisions robots and journalists working together, it describes automation as enabling news organizations to expand their coverage by freeing up reporters to focus on more challenging topics and issues.
So far, for example, automated business news has vastly increased the number of publicly traded companies receiving coverage since 2014, while flagging anomalies or other potentially newsworthy findings for journalists to pursue further.
In a report last year, researchers at the University of Washington and Stanford University found something interesting about all that automation: The volume of shares being traded in companies that previously received little or no news coverage had measurably increased.
Last year, for the first time, recaps of every Minor League Baseball game were produced by robots for The Associated Press, something the wire service has explained never could have been accomplished with its existing staff.
So, where does that leave us old-fashioned writers?
In a head-to-head showdown between a news-writing robot created by Automated Insights, the company that developed the algorithms used by The Associated Press, and National Public Radio’s White House correspondent Scott Horsley, the difference in quality between the two versions was obvious.
But the algorithm was much faster.
Both had been given the same data, a first quarter 2015 earnings report for the restaurant chain Denny’s, and immediately went to work.
Horsley, a former business writer, produced a longer, easier-to-read report that showed nice deadline creativity. Here, for example, was his lead paragraph:
Denny’s Corporation notched a grand slam of its own in the first quarter, earning a better-than-expected 10 cents a share, as restaurant sales jumped by more than 7-percent.
The algorithm, meanwhile, accurately identified the most important elements within the data, though it didn’t mention that the earnings beat expectations until the third sentence, and took a much more basic approach to the writing. Here’s how the robot, called Wordsmith, opened the article:
Denny’s Corporation on Monday reported first-quarter profit of 8.5 million dollars.
Simple and unimaginative. But it finished the entire article in under two minutes, and when it comes to boilerplate type news like this, it’s arguable how much attention the average reader gives to the overall quality of the writing.
That’s a tough realization for any writer.
Similar comparisons could be drawn in evaluating human versus robotic sports coverage.
The giveaway clue that a game recap was written by an actual human typically involves the use of details you won’t find on standard statistical spreadsheets. Did a team pull off an upset win even without its top running back, who is out for two weeks with a torn ACL? Did the winning 3-pointer at the buzzer come from a half court desperation throw?
But, again, it’s debatable whether the average reader is looking for that level of detail in every article that shows up in a news feed.
We’ll keep trying, though.
The Associated Press has pledged to use AI only to expand its coverage options rather than replace journalists. Its experiments with automation have focused on bringing narrative-style coverage to lower-division collegiate leagues and teams it otherwise lacks the staff to cover in person.
Skepticism over how long that commitment can stand, or if other news organizations will adopt the same approach, is high among many.
But with the news industry struggling financially as advertisers flee for cheaper online platforms, perhaps automation might be one of the keys to keeping independent journalism viable as technology steadily transforms industries of all sorts in the coming years.