Burriss: Robots now writing news
Tuesday, June 9, 2015 12:02 pm
By LARRY BURRISS
We've talked here a few times about people fearing robots, or androids, or cyborgs, or whatever they are called, are becoming more and more human-like. I've usually dismissed this kind of talk as being either silly or hyper-alarmist. That is, until I read a couple of stories about how some news stories are now being written.
Apparently there is software that, given a set of parameters, can generate well-written news stories automatically.
For example, The Los Angeles Times has an application that uses information from the U.S. Geological Survey to generate stories about local earthquakes. The cyber-stories tell where the earthquake was centered, the magnitude, the time, and the nearest cities and towns.
The system can also generate background information such as the number of recent earthquakes of a given magnitude.
Now, we all know computers know where you are located. But suppose you are reading a story that has some statistical data on The New York Times web site? Without even asking where you live, the newspaper computers can easily generate a local angle to the story in real time.
Not only will it write paragraphs with the local data, but it will also compute comparisons between your specific location and similar locations. And not just with the raw numbers, but with-written commentary as well.
Why can't you just enter your location yourself? Because research has shown when users are asked to enter their ZIP code, readership drops significantly. So the system enters your location automatically.
Where can this kind of data manipulation take us? Well, I bet it's not a far stretch to combine real-time data analysis with facial recognition software that can be programmed to recognize emotions.
So a political candidate could, in the middle of a speech, have the teleprompter change the speech on-the-fly to accommodate shifting audience moods. Which could be just one more example of us not knowing what a candidate really thinks, but rather, what the data manipulators think the candidates should say they believe.