Sirius Says "Hello" – The FOSS Digital Assistant

Sirius is the FOSS alternative to Siri, Cortana, and Google Now

Sirius is a project from the University of Michigan that I believe is going to help drive innovation in the Intelligent Personal Assistant (IPA) space. The technology behind these systems is brilliant and compelling. The AI being developed for the applications is astonishing and while it’s not quite at the level of conversational dialogue that science fiction has long offered up, it has certainly come a long way. Until now however, I’ve had serious misgivings about using IPA systems. So what issues are there which are worth getting all worked up about?

Sirius - the Open Intelligent Personal Assistant
Johann Hauswald of the University of Michigan working on Sirius code.

 I know what you searched for last summer.

Terrible 1997 slasher movie reference aside, the fact remains that companies retain and analyze the queries their users make via their IPAs. The problem with this is multifaceted from a privacy standpoint.

First of all is the fact that your data is only protected by the privacy policy at the company in question. Can they share your data with third-party entities? Can they sell your data? Do you even know? If they change their policy would you even notice? Most privacy policies are “subject to change without notice.”

So what do the companies do with the data collected from millions of daily searches performed on their platforms? From a software design standpoint they obviously want to review the metrics of successful searches and so forth to determine platform enhancements. Beyond simple software feedback, they can already use it for targeted advertising, for one. Apple has recently been “outed” by an employee as sharing the data with third parties.

“Guys, I’m telling you, if you’ve said it to your phone, it’s been recorded… and there’s a damn good chance a third-party is going to hear it,”

So, there’s that.

Sirius is set to change this paradigm

In a culture where there it is customary to pay for convenience with our privacy, Sirius is helping set the stage for a revolution in IPA. Because Sirius is completely open source it will enable other vendors to adopt and adapt the technology for both private and commercial uses. There are however a few interesting uses that can be immediately derived.

  1. Truly Personal Digital Assistants – not cloud based. Your IPA, your computer (or phone) and your data.
  2. Search services such as Startpage may offer IPA-based search services while sanitizing your private data.
  3. Companies, universities and other institutions could have custom organization-based databases that your IPA would be able to tap into. For example you might visit a historical site with your family. When the inevitable questions from the kiddos arise you could connect your IPA to the organization’s database and ask away. There is opportunity for a learning AI to ingest that information to have access to it at a later date as well.

For a truly personal IPA you need to store your own data – so a personal cloud would come in handy. Perhaps there will be plugins for ownCloud (as detailed previously) or other similar projects to store and access data directly from your private cloud.

Once mobile computing storage density is such that the data and be stored and analyzed on-device, the cloud based portion of the system may be simply for data redundancy.

The information is inside the computer!?

Sirius - Inside the Computer!?

Sirius is already capable of voice, text and image searches and currently utilizes Wikipedia for data.

Sirius query flowThe open source nature of Sirius will enable individuals, research teams, universities and companies to advance the machine learning aspects of the technology and improve the data processing, machine learning, and data storage mechanisms of the software. This is very much in the same vein as the GNU/Linux development model which has driven a lot of innovation in the tech sector for the last 20+ years.

The future is now… and later, too.

I am, admittedly, a little fan-boyishly exited over Sirius and the core concept of what the team at the University of Michigan has achieved with their project. The fact that this has been so willingly open-sourced can only help drive innovation in this sector. I’m a realist in the understanding that the code will be forked and used for purposes other than those that I’ve outlined above, but the fact remains that this is going to enable a slew of developers to progress in making IPAs an increasingly useful tool.

With all this talk of IPAs, I’m suddenly thirsty… time to compile Sirius on an Ubuntu box and ask for some recommendations for a good cold beverage. Just imagine, I could ask my IPA about IPAs; it’s IPA-ception!

While we may not be living in an episode of “The Jetsons” just yet – the arena of machine learning is upon us in full force and I’m glad that in Sirius an open source addition has been added to the mix.

If you’d like to read the more technical details about Sirius you can find the research documentation here.

Leave a comment

Your email address will not be published. Required fields are marked *