Yury Millner is the chap in question here, a famously media-shy person who rarely grants interviews. So his recent words with the Russian business daily newspaper Vedomosti are worth paying attention to, not only because one of his companies (DST Global Investment) has a big minority stake in Facebook, but because he’s chairman of Mail.ru, which dominates the massive Russian email scene, and which recently raised around a billion dollars in an IPO.
So what did Millner say? Among other pronouncements, he remarked: “I think that in 10 years if you ask a question on a social network and you get an answer you will not know if a computer or a person has answered you.” It’ll work the other way around too, Millner thinks: “When you receive a question, you will not know if it has been asked by a person or an artificial intelligence. And by answering you help the computer create an algorithm.”
This is a pretty bold suggestion, but it’s backed up by some solid science. There’s already an app on Facebook called Ultra Hal (inspired by the computer in 2001) that is a web-interface to an artificially intelligent chat interface. Made by Zabaware, it won the 17th annual Loebner Prize for AI–based on the Turing Test for how realistically convincing an AI is. Hal lets Facebookers chat to it, and it actively learns to improve its intelligence during the discussions. Zabaware sells a commercial version that’s smart enough to be “used as a companion or entertainment product” and “can discuss any topic” or “be used as a personal or office assistant.” Back in November, Spanish scientists crafted a program which can recognize emotions in a human voice–vital for well realized AI in the future, and influential futurist Ray Kurzweil has a big money bet that a computer will ace the Turing Test by 2029.
AI is booming at the moment, and Facebook (and its contemporaries, like Twitter) is an excellent platform on which to test AI–there is a wealth of conversational data in all languages, perfectly suited to honing an AI’s talents. Plus AI could actually find a place inside Facebook’s workings–as the site gets bigger and more influential, and more people use it to share more information, finding what you prefer to read inside your Facebook page will be difficult. An AI system could be quite helpful in sifting through reams of data, as it could be programmed to learn what you prefer, and even to guess at things you may be interested in outside of your normal habits. Is this what Millner is hinting at?
To read more news on this, and similar stuff, keep up with my updates by following me, Kit Eaton, on Twitter.