Virtual, personal assistants seem to be the rave these days on mobile, from the big ones like Siri, Cortana, Google Now, and most recently BlackBerry Assistant, to the little known apps and services scattered throughout app markets. So it isn’t surprising that we’re hearing about another one called Sirius, a not so subtle play on Siri perhaps, but this software, and we can’t call it product yet, hailing from the University of Michigan is a bit different. For one, it is open source software. And quite surprisingly, it has the financial support of Google.
Sirius isn’t shy about acknowledging its likeness to those big three (or four) virtual assistants, so it is a bit interesting that Google would support it. Then again, it may actually even benefit from it in the long run, given Sirius’ open source nature. Also, being more of a research project than an actual product, Sirius isn’t exactly a direct competitor to Google Now, much less a commercial rival.
While Sirius does have speech recognition like the others, it does have a few unique features. For example, it also has image recognition, which would allow users to feed it an image or take a photo from their smartphone and ask Sirius questions about the photo. It sounds somewhat like Amazon’s Firefly service, but one that’s not limited to simply recognizing an object but also deriving related information about it based on user questions. It also features text recognition so that it can crawl through, for example, Wikipedia entries in search for answers.
But Sirius’ real character is its openness. Research project head Jason Mars compares it to how Linux is the open source Windows, so to speak. The bottom line is that users and developers will be able to take Sirius apart and learn from it, improve it, and customize it to their own needs, something you will never be able to do with Siri, Cortana, or Google Now. Especially not at source code level. Sirius itself stands on the shoulder of giants, utilizing other open source software for its own functionality. Image recognition, for example, is OpenCV (Open Source Computer Vision), while its Question and Answer system utilizes OpenEphyra. Unlike Siri and the others, however, Sirius so far has been tested to work on the Ubuntu Linux desktop only, but could soon expand to other operating systems as well as mobile platforms.
The promise of an extensible, customizable, and open source intelligent virtual assistant seems to appeal to more than just Google. DARPA as well as the National Science Foundation seem to also be on board. The source code, as well as instructions to build Sirius, can be found on Github.