IBM is working to turn its Jeopardy-winning supercomputer into the newest mobile device personal assistant.
Dubbed "Watson 2.0," the system would be a voice-activated "supercharged" version of Siri, at least that's how Bernie Meyerson, VP of Innovation at IBM, described it to Bloomberg this week. We all know Watson is super smart - there are almost a dozen racks of IBM servers in New York that can back up that claim - but is he helpful?
Adding the ability to sense and assess real-world input like image recognition, location data and voice recognition are a few of the tools needed to make Watson into something more than a the system more user friendly. To begin with, IBM will have to include some of its existing technologies, like image interpretation, into the system. It might also have to license voice recognition technology from a company like Nuance, much like Apple did with Siri.
Despite the obvious similarities it doesn’t seem like IBM has plans to take down Siri; it wants to attract a different kind of user base: the corporate world, more specifically, finance and healthcare.
Last year, IBM teamed up with Nuance to research the possibility of using Watson's brains and Nuance's voice recognition technology to create a smart, fast personal assistant and fact-finding machine for doctors and nurses. In order to do that, Watson had to "learn" everything there is to know about a multitude of different medical topics.
Last September, Watson began studying oncology through a partnership with health insurer, WellPoint. Researchers "taught" Watson in a backwards kind of way, feeding it answers to questions they had developed. When a question is posed, Watson uses information it has already learned, then accesses millions of books and websites at 66 million pages a second, and then answers.
The more it analyzes and is told which answers are right, the more accurate it becomes. Watson will not only tell you the answer, it'll show you why it's is right. This is promising, but it's already been a year and Watson isn't projected to be an oncology expert until 2013.
Another problem is that Watson is too smart for a device's battery life. Watson uses the power equivalent of 6,000 desktop computers. That makes it too energy consuming to be a viable smart phone application. Meyerson says that the mobile version they are working on uses less power, saying, "The power it takes to make Watson work is dropping down like a stone." Addressing that issue could turn Watson into a leaner assistant to be used on any device.
For now, "Watson 2.0" is still in testing phases, but it's on its way to moving from a server stack to something you can hold in your hands.