When an earthquake hit Los Angeles recently, Ken Schwencke, a journalist and programmer for the Los Angeles Times, was first to get the news out. Woken up by the tremors at 6:25 a.m. on Monday, March 17, he went to his computer and found a brief story already waiting, courtesy of a robot — an algorithm he developed and named Quakebot.
Quakebot’s role in the swift reporting of the earthquake story has industry observers talking about the role of robots in the future of journalism. Among those at the forefront of robot journalism is Noam Lemelshtrich Latar, dean of the Sammy Ofer School of Communications at the Interdisciplinary Center in Herzliya, Israel. Latar has written several papers on the topic, such as “The Future of Journalism: Artificial Intelligence and Digital Identities” and “Digital Identities and Journalism Content: How Artificial Intelligence and Journalism May Co-develop and Why Society Should Care.”
Latar has a master’s degree in engineering systems from Stanford and a Ph.D. in communications from MIT. His work has been concentrated in the area of touch-screen phones and allied devices. His paper, “Screen Feedback from Home Terminals,” was the first to explore this concept. Today, however, Latar is focusing on artificial intelligence and robot journalism. In this interview with Knowledge at Wharton conducted late last year, Latar discusses whether robots will one day replace human journalists.
An edited transcript of the conversation follows.
Knowledge at Wharton: What exactly is robot journalism?
Noam Latar: Computers have helped journalists to write, to find facts, since the middle of the last century. There was what we call data mining and analytics — data analytics — which helped journalists find the facts and do investigative journalism. So, this is not new. What is now developing is that the new programs — artificial intelligence (AI) programs — get the facts and write the story within a fraction of a second. Today, there are stories written in Forbes and other newsmagazines that are untouched by human journalists. The AI program writes the story, and the name of the journalist is really the name of a robot. There’s a company called Narrative Science in Illinois that is already doing it and has collected a lot of money from investors.
Knowledge at Wharton: Is this a new phenomenon, or has this been around for a while? Didn’t your research at MIT in the 1970s actually predict some of this?
Latar: No. My research predicted touch screens, which were later used by Steve Jobs. I did the first studies on interactive television. We studied how providing people in the television studio with devices to provide feedback would affect the group dynamics and the discussion dynamics. At that time, we did not predict robot journalism at all. Data mining has developed in the past 20 years.
“I foresee that the new leader of the newsroom will not be the experienced journalist … but the computer engineer.”
Knowledge at Wharton: How pervasive is robot journalism?
Latar: It’s still in the initial stages because the programs really started in 2010. But they’re penetrating very quickly because the robot journalist has certain real advantages. First of all, it never forgets facts. It can do research very quickly. It never asks for a day off. And it can write the story within seconds. If you write the program correctly, [the robot is] not even biased. As you know, most journalists are biased about their stories. But the robot journalist, if you program it correctly, can be completely unbiased.
Knowledge at Wharton: So they don’t miss deadlines?
Latar: For example, there was an earthquake in California [last year] and a robot journalist wrote the story; the journalist under whose name the story appeared was already asleep. The Los Angeles Times has been using this kind of robotic story writer for quite a while.
Knowledge at Wharton: How fast is it growing, and what are the factors driving the growth?
Latar: That is a good question. The ecosystem that really promotes this robotic journalism is, of course, the multimedia platform and the slow disappearance of print. There is the question of cost. Advertising money is today shifting from print to the Internet. And because robotic journalists are cheap, efficient and quick and because of economic considerations, I expect that robotic journalism will grow very fast.
The optimists view this entry of robotic journalism as a new era where really good journalists will not disappear, but will be forced to think again about how they can be innovative, do more in-depth analysis. So, it depends how you see the cup — half full or half empty. I think journalism will improve; the threat of robotic journalism will always exist. Any newspaper — electronic or print — that you purchase in the coming years will have a major part of the stories written by robotic journalists. But people will still seek a human journalist because they have always added value, added innovation and added perspective. So, I see this as a positive force in improving future journalism.
Knowledge at Wharton: Since you mentioned costs, I wonder what sort of a business model can be put around this kind of journalism?
Latar: I foresee that the new leader of the newsroom will not be the experienced journalist … but the computer engineer. Just the other day, [Arthur Ochs] Sulzberger — the publisher of The New York Times — was asked: “What would you do in the light of this digital development in media?” And he said: “I would hire more engineers.” So I see tremendous saving in costs in future newsrooms that are going to be fully automatic. The leaders will be those journalist geeks who understand the media and know how to use data mining and how to add value to what the robot journalists can do. But it’s going to be very efficient.
Another aspect of it — today we are going into targeted advertising. The robot journalist will be able not only to write a story but also to immediately send it to you if it knows that you are interested in a certain type of information. So, we have here a complete automation of news gathering, news analysis, story writing and the targeting of the information.
“People will still seek a human journalist because they have always added value, added innovation and added perspective.”
Knowledge at Wharton: That’s fascinating. To what extent do social media platforms, like Facebook, Twitter or LinkedIn, factor into this phenomenon?
Latar: Social media and social networks help spread the word of all these new developments. Once our good, efficient robotic journalist is in operation, these social networks — especially Twitter — will spread the word very quickly. It will expedite the development of these processes. I’m more of a pessimist about Facebook because I think it will probably not change the way it is today; it will probably disappear in favor of the Twitter kind of systems, which will also develop into video. I haven’t looked at my Facebook page in the past two weeks because it’s loaded with information, loaded with dogs and friends and everybody’s happy traveling all over Europe and America. I’m the only one who’s working.
I think the present structure of Facebook will have to change. But Twitter is [going] in the right direction — quick, to the point. So, social networks will spread the word of robotic journalism very quickly.
Knowledge at Wharton: How much of robot journalism do you see in Israeli media?
Latar: Not much, because … the software to write the stories and to do the data mining is quite expensive. Israel is more in the direction of developing start-ups that point the road to this new software and new AI techniques. The start-ups are sold to American or European companies. Israel is more [focused on] developing the seeds of these innovations. So far, Israel has not been very successful in taking small ideas and making them large scale.
The media in Israel is in a very sad situation. It is controlled by the government, by politics, by rich people, by capital. The Israeli media today is nothing to write home about. It’s quite sad.
Knowledge at Wharton: I’m going to ask two or three questions wearing the hat of a human journalist.
Latar: I’ll try to answer also as a human professor.
Knowledge at Wharton: A lot of successful journalists share three qualities. The first is they have relentless curiosity. The second is they have innate skepticism, especially about PR. And the third is that they have the ability to tell stories simply and clearly in a way that they’ll be understood by a large number of people. Do you think these are functions that can be automated?
Latar: Data mining and software that does data mining in huge data silos … is really the ultimate expression of curiosity. In my article, I talk about this young lady from Israel who was declared by MIT as one of the [top] future scientists. She discovered that if, in a rural area, a poor area, you have drought one year and a flood the following year, then in the third year there will probably be an outbreak of cholera. She accomplished that by looking over a huge amount of data. All human journalists cannot have this ability to scan huge amounts of data.
Knowledge at Wharton: But it was a human researcher who found it out — who made those connections.
Latar: She wrote the program. Now this program can go to other data silos and discover other things. The robot journalists will force human journalists to be on their toes and be more curious. Skepticism is one value which can be translated to statistics. A good program can be skeptical about a result. But I don’t see good journalists disappearing. I see fewer journalists but better ones in the light of robotic journalism.
“I think the present structure of Facebook will have to change. But Twitter is [going] in the right direction — quick, to the point.”
Most journalists are quite lazy. You have to provide them with a story. You have to provide them the information. And they publish it under their name. And they don’t even go out to do investigative journalism. Investigative journalism is risky. It’s costly. It’s dangerous. And fewer and fewer people are doing it. Robotic journalism, I hope, will bring this to light again.
Knowledge at Wharton: The other quality — and this is more for editors than for reporters — is that editors exercise editorial judgment. I wonder how robot journalists fit into that category?
Latar: [Look at] Google News. Google News crawls over 27,000 news sources today. They wrote a program [called] “How to Exercise Judgment.” They take an event, for example, and they can study what is the probability of this event occurring. They bring the criteria that every good editor exercises and put it into their programs. It’s like a good chess player; a good chess program will never improve itself by gaming against a weak partner.
The people who write these programs go to the best editors. They follow their best practices and interview them. And they try to write a program to ask the same questions that a good editor would ask.
Knowledge at Wharton: In conclusion, let’s end with the question we asked at the beginning. Can robot journalists replace human journalists?
Latar: Yes, but not completely. But there is one pessimistic ending to this. Robotic journalism from the point of view of economics is very efficient. And a very good journalist can be quite expensive. I fear that economic considerations in the news media could force more and more media organization to go into robotics because of the cost advantages. As a human being, I hope that at least a certain portion of every media in the future will be run by high-quality journalists who will make sure that the human journalist is always better than the robot journalist. That is my hope.