Forget Malcolm Gladwell and Tom Friedman. The next great journalist could be a computer program.
Already, artificial intelligence has been put to work at Forbes, the Associated Press, Reuters, and The Big Ten Network. Coverage of NCAA basketball games, earnings reports, and dips in housing prices are regularly penned by this new breed of “journalist.”
And according to Narrative Science’s Kris Hammond, computers are only going to get more creative. Hammond, who’s also a professor of computer science and journalism at Northwestern, believes that artificial intelligence systems can learn to break away from restrictive formulas and absorb subtle linguistic idiosyncrasies. I spoke with him about where he sees AI going.
[This interview has been edited and condensed. For the full conversation, visit innovationhub.org.]
Kara Miller: What will we be able to program a computer to write in the not-too-distant future?
Kris Hammond: Any place where we see data—numbers and symbols—there will now be a possibility of a story. And there is data everywhere. If you wear a Fitbit, there’s data. If you have a Nest in your home, there’s data. If you drive a car, there’s data. But we are living in a world where people have to go to the machine to get data about what’s happening. And they have to figure it out and fight with it. So at Narrative Science, we are building a world where that fighting goes away, and where the machine’s job is to explain what is happening in the world on the basis of the data that it has.
KM: How do you teach an AI system about things like euphemisms, irony, and slang?
KH: You can teach anyone or anything. To teach a system to speak ironically or to generate texts in an ironic form, you have to teach it that it’s another way to express a strong negative opinion. The system needs to learn to amplify the positive so far that it becomes absurd. For example, you have a company that explodes and crashes in record time. And you can say that Company X was the success story of an era. A system will learn irony in the same way that we learn irony, except maybe ever so slightly faster.
KM: Can AI programs break free from things connected to data and write creative texts like screenplays and television shows?
KH: This is something we have looked at on the laboratory side at Northwestern. The question is, can you look at the nature of story and turn it into something that the machine can understand? And I would argue, absolutely. In fact, if you take a look at any screenwriting book, the notion is that movies have the same formulaic feel to them. So if you can identify a pattern and think of how to build up the components and how those components relate to one another, then certainly you can have a machine write creative texts.
KM: Do you think we overestimate the power of creativity?
KH: I think creativity is magical. But, it’s something that can be learned, it can be taught, and it can be trained. And at the very bottom, it’s going to be something stupid. Intelligence is built on a whole bunch of stupid, but then it becomes intelligent. For example, if I teach you the twenty rules for being an umpire at a baseball game, then you can be a great umpire. But it’s based on simple rules. Likewise in creativity, you can teach a machine to be creative.
KM: What is it going to be like when we go to museums and computers have produced our art, or we turn on the television to watch a show that a computer wrote? Does that worry you at all?
KH: I think it’s phenomenal that we will give machines the gift of intelligence, creativity, and awareness. I think building AI systems is like having kids. You want your kids to be smarter than you, you want your kids to be more successful than you, and you want them to be more creative than you. It’s not going to be a horrible world; it’s going to be a magnificent world. Because it doesn’t diminish us to be surrounded by things that are smarter than us. I think we can do better because of machines—we can have a richer world.
Tricia Breton contributed to this write-up.