By popular demand (though without confirmation from Microsoft), the fictional artificially intelligent being that first appeared in the xBox videogame “HALO: Combat Evolved” and its sequels, is being recruited to be the “Siri-equivalent” for Windows-based phones and tablets. As indicated by this post from June of 2013 in Microsoft’s Windows Phone Community page, the idea of having Cortana as an Intelligent Virtual Assistant has been a hot topic among Microsoft’s fan base for a while. Today, coverage by Tom Warren in The Verge, based on a “leaked” SDK, shows that the Bing search box which has graced past versions of Windows Phone 8, will be replaced by the ever-conversational Cortana, who will have a personality “like Siri” but bring some of the predictive assistance functions that are popularizing Google Now. (Warren’s article sparked coverage of the imminent launch by Eric Limer in Gizmodo and by Darrell Etherington in TechCrunch.)
Perhaps most import is that Microsoft sees Cortana on a smartphone as the natural language-based controller for apps and media that run or play on a Notebook. According to Warren, Microsoft will position this approach as a form of privacy protection. Over time, Cortana will learn a lot about her users’ activities, likes, dislikes, preferences and other personal data. It will also be set up to react to the content of emails and alerts automatically, including the management of “Do Not Disturb” functionality that will let an inner circle of friends and family get their messages through during quiet hours.
Microsoft has been totally mum on the topic, so it is unclear whether it will be using its own speech recognition and text-to-speech software. Ford Sync, for instance, which is “powered by Microsoft” is an amalgam of speech processing resources, including Microsoft’s home-grown speech recognition and Nuance’s text-to-speech. As The Verge’s Warren notes, the xBox voice command is highly accurate, but appears to have a relatively small vocabulary (mostly commands). To be “Siri-like” Cortana will have to be accurate, human-sounding and possessed of a relatively large vocabulary, or at least one that can speak to individual users in their own terms.