This Wasn??t Written by an Algorithm, But More and More Is
Source: Timothy Aeppel
Scott Frederick, chief operating officer of Automated Insights, displays some of the automated statistics and news provided online by his company in 2012. The company was founded by a former Cisco engineer and started, like Narrative Science, with sports stories.
Robots have conquered assembly lines. Now they’re going after the wordsmiths.
In a downtown Chicago office tower, a team of scientists are at the cutting edge of automating a task until now only done by the likes of journalists, bankers, insurance salesmen and government spies: Writing up thoughts and observations in language that reads like it came from a human.
Narrative Science was founded by researchers from Northwestern University, who discovered they could use computer algorithms to generate basic sports stories. They still do that, but have added dozens more clients, including financial services companies, retailers and government intelligence agencies.
They’re not the only ones doing this. Automated Insights, in Durham, N.C., was founded by a former Cisco engineer and also started with sports stories. Some companies are innovating on their own. The Los Angeles Times developed a computer program that writes barebones versions of stories about earthquakes and local crime, pulling information directly from government websites.
The trend is part of a larger thrust of automation into the workplace that’s changing how people do some white-collar jobs, which constitute eight out of 10 jobs in the economy. Computers now do legal research, translate conversations and help diagnose diseases.
In some cases, the new technology is augmenting human talent rather than replacing it. For instance, one Narrative Science client is Credit Suisse, which uses algorithms to generate summaries of stock activity used inside the bank by brokers and others who need to figure out what’s going on when share prices fluctuate.
That task was formerly done by 20 employees scattered around the world. It was time-consuming, which meant they could only track about 1,500 companies. And since not all were native English speakers or equally experienced, the quality varied.
The new technology allows Credit Suisse to cover 5,000 companies, because the bulk of the work is automated. They didn’t shed the 20 workers. Instead, those workers focus on writing more detailed reports on only the largest companies. In theory, they would have had to vastly expand the staff to cover 5,000 companies―so that’s job growth that won’t happen because of the technology.
One fast-food chain uses the system to scan point-of-sale information for 14,000 franchises and generates a weekly report tailored to each location. “They say things like, ‘Based on the data, if you sell six more chicken sandwiches, you’ll make X more a week, and Y more a year,’ ” says Kristian Hammond, the company’s chief scientist and a computer science professor at Northwestern.
Another example is at car dealers. Narrative Science allows them to study inventory lists and generate advertisements to sell cars on auto websites. The tool is designed to highlight what makes a car attractive―knowing, for instance, when to throw in details about the low mileage or style.
“It takes my thinking process out of it,” says Bubba Dickson, general sales manager at Tom Light Chevrolet in Bryan, Texas, who estimates he used to spend several hours every other day on the task. “This thing is a lot better at knowing what to say to pull someone in, and it frees me up to other things.”
| }
|