Breaking News: Robot Reports An Earthquake
SCOTT SIMON, HOST:
This week, an earthquake gave an algorithm its big break in journalism. The 4.4 earthquake rumbled the ground of Southern California on Monday morning, and within three minutes, Quakebot generated an article about it for the Los Angeles Times. And Quakebot probably didn't charge overtime. Ken Schwencke joins us now. He is the journalist and programmer who designed Quakebot. Mr. Schwencke, thanks so much for being with us.
KEN SCHWENCKE: Thank you.
SIMON: So, the article is essentially prewritten or I'm trying to figure this out.
SCHWENCKE: It's basically Mad Libs - Mad Libs for Earthquakes.
SIMON: I haven't played that game. So, you guys get an alert from the U.S. Geological Survey there's an earthquake happening and...
SCHWENCKE: And then basically they send an email to us - anybody can subscribe to these emails - and we take out the information from the email, which is like a big list that says, you know, it was a four-point - you know, at the first blush, they said it was a 4.7, five miles away from Westwood. And so I have a fill-in-the blanks template that it then puts in, you know, 4.7, the word shallow, based off of the depth that the USGS says it was, and fills in the other basic pieces of information, and then fires that into our website.
SIMON: Who got the byline?
SCHWENCKE: I did.
SIMON: You did?
SCHWENCKE: I did.
SIMON: Not the algorithm?
SCHWENCKE: Not the algorithm. He gets a little credit at the bottom.
SIMON: How much of what I read every day in the L.A. Times is written by an algorithm?
SCHWENCKE: Well, the only places we really use it is for, you know, earthquakes, and just the initial reports, and if you are a reader of our homicide report, which tracks every homicide back since 2007, anything that we couldn't get further information on, that we just have basic details on, gets templatized into a robo-post as well.
SIMON: Do you think there's going to be more of us in what we call journalism in the future?
SCHWENCKE: I think, you know, there are a lot of companies and people experimenting with it, especially on, you know, much higher levels than I'm currently implementing. But at least here, you know, these are just - we're kind of using them for supplemental, like, starting information and then we tend to fill them out after the fact.
SIMON: There's really only one of determining whether or not this algorithm's up to the job of being a reporter. Does the algorithm complain about its editor?
SCHWENCKE: You know, he doesn't.
SIMON: Well, he's not a reporter then - he or she or it. Yes.
SCHWENCKE: He's actually very friendly.
SIMON: Very friendly, really, and thanks the editor for improving his or her work?
SCHWENCKE: He does. He even asks for changes.
SIMON: This is sounding awfully inauthentic to me.
SCHWENCKE: Well, I figured if I could program the personality I might as well make it friendly.
SIMON: Ken Schwencke of the Los Angeles Times. Thanks so much for being with us.
SCHWENCKE: Thank you.
(SOUNDBITE OF SONG, "SHAKE, RATTLE AND ROLL")
SIMON: This is NPR News.
NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.