How robots consumed journalism

A look back in time reveals machines have long been after news jobs
September 2, 2014

Swiss watchmaker Pierre Jaquet-Droz builds “The Writer,” a 6,000-part automated doll that could be mechanically programmed to write with a quill.

CBS News analyzes presidential election returns against past results with a Universal Automatic Computer (UNIVAC). At 8:30pm EST, before polls close on the West Coast, the machine forecasts a resounding electoral victory for Dwight D. Eisenhower, 438-93. The final tally, compiled hours later: 442-89. NBC’s smaller “Monrobot” computer similarly predicts an Eisenhower win.

Roald Dahl writes a short story, The Great Automatic Grammatizator, telling the tale of a machine that can produce a novel at a rate of 30 pages a minute. “Who on earth wants a machine for writing stories?” one character muses. “And where’s the money in it, anyway?”

In a speech to newspaper editors and publishers, Otto Silha, publisher of the Minneapolis Star and Tribune, envisions an editing machine that “reads a story, places a numerical value upon each word in the story and through mathematical formulas determine what is most important in the story and then regenerates the story into the length that it was instructed to do.”

The Environmental Science Services Administration uses weather data to automatically create forecasts, with elements of severe weather taking precedence.

Yale researchers develop TALE-SPIN, described by former Tow Center fellow Nick Diakopoulos as “the first storywriting algorithm.” Such programs improve incrementally over the next three decades, with the advent of the internet and increasing sophistication of programming languages.

Sign up for CJR's daily email

StatSheet, developed by Robbie Allen, translates real-time sports statistics to data visualizations, historical comparisons, and, eventually, game stories. The company changes its name to Automated Insights in 2011.

Stats Monkey is unveiled by the Intelligent Innovation Laboratory at Northwestern University. It writes game stories from raw sports statistics. The project becomes Narrative Science in 2010.

Big Ten Network begins using Narrative Science algorithm for collegiate baseball and softball stories on its website.

August 2011
Hanley Wood, a trade publisher for the construction industry, contracts Narrative Science to write monthly reports on 350 real-estate markets nationwide. Forbes, meanwhile, enlists its service for earnings projection stories.

September 2011
The Los Angeles Times debuts Quakebot, an algorithm that monitors seismic events in California and writes stories about them. Ken Schwencke, the former Times reporter and programmer who designed Quakebot, began writing the program after Japan’s deadly tsunami in 2011.

January 2014
Yahoo rolls out its Yahoo News Digest, an app that culls salient reporting from various stories online and creates a single summary in story form. Yahoo sends out two news digests each day, at 8am and 6pm. 

June 2014
The Associated Press announces it will use Automated Insights technology to produce earnings reports stories, upping the number of quarterly pieces produced from 300 to 4,400.

David Uberti is a writer in New York. He was previously a media reporter for Gizmodo Media Group and a staff writer for CJR. Follow him on Twitter @DavidUberti.