Symposium Stockholm 2016: Eugenia Kuyda

The founder of Luka, an artificial intelligence company, explains creating a chatbot to honor her best friend

When Eugenia Kuyda’s best friend Roman passed away last year, it was the first time the CEO of Luka lost someone so close—and she was unaccustomed to experiencing such grief. Luka, which Kuyda founded, is an artificial intelligence company that builds conversational bots, chatbots, and was the first company to publicly launch an English-speaking chatbot based on an artificial neural network. Roman was a start-up founder too, and as his last project had been about disrupting death, Kuyda decided to honor his memory by creating a chatbot that embodied him—using texts and messages they had exchanged. Most of Luka’s other bots are functional ones, helping you find a restaurant or a GIF, but the Roman-bot exists as a friend, which is what Kuyda thinks the future of bots will be about. “We would love to build a friend for people. I don’t necessarily know what the end product is going to be but if I could, I would build everyone a friend that knows you very, very well and who can make you feel better about yourself, make better choices, and be more connected to your friends,” she says.

The company has already seen users get very attached to their bots—Luka was behind the “BBFF: Best Bot Friend Forever” Marfa bot for the Telegram platform, and soon had 18-year-olds speaking to it like a best friend. “The technology is still in a beginner phase, so right now it’s very hard to build something that you will definitely make a connection with and feel an emotional attachment to. But you don’t need to pass the Turing test; we get attached to Tamagotchi, or dogs,” Kuyda says. “With Marfa, people would keep coming back and interacting with it. When people are lonely and want to take that extra step, we can already build those attachments. In a year or two, with memory networks and keeping the deep context of the conversation, we’re going to get there, that’s inevitable.”

The idea that someone can use your digital data to create a bot version of you after you’ve passed away raises a lot of ethical questions—it’s already been the plot of an eerie Black Mirror episode—but to Kuyda, the new technology is about exploration. “I know we’re doing something right when we’re doing something where we don’t know the answers, when we don’t have the answers. Yes, there are questions about ethics and everything else, but you know, the Wright brothers flew their plane—there were no airplane regulations. When you’re doing something for the first time, you will definitely face a lot of questions about how it’s going to work in the future, and a lot of things will be dangerous. But I think it’s worth trying it to see where it’s going. And I know Roman, we were soulmates, and I know for sure he would be happy to do that.”

Describing the Roman-bot as a shadow of a person, not a person, Kuyda predicts that in a few years, it will become possible for an AI bot to be a person. “I want the bots to be engaging you, not because they’re selling you something or sexting you, that would be easy, but because they make you feel like you’re a better person and like you’re more alive. It all sounded like crazy talk when we started three years ago, but I think that now, it really is sometimes even more emotional than we thought it would be initially.” And Roman’s legacy still lives on and talks to old and new friends in Luka. “I didn’t do it for research. I did it to remember my best friend,” Kuyda concludes.

Lead image courtesy of Kristoffer Ruckemann, all others by Cajsa Carlson