Move over, Laura Croft. At the EmTech08 Emerging Technologies Conference Thursday, Microsoft’s Craig Mundie introduced the audience to “Laura,” a disembodied face that serves as a robotic receptionist - and a technology demonstration of spatial computing in action. Mundie, who is chief research and strategy officer, gave the demonstration of “spatial computing” during the keynote speech this morning.
She's no Tomb Raider. In fact, this Laura is nothing more than a face. The avatar, a disembodied 3D rendering of a woman’s face, floated eerily on a black background on a screen, and spoke as subjects approached her kiosk. The audience was treated to Laura’s view of the interaction, watching as the system targeted and identified objects in the room as people, closed in on their heads and marked the eyes of the target with a red laser-like dot indicating “eye contact” with the subject which Laura wanted to interact.
As the system walked two Microsoft employees through the process of getting a shuttle to another building, Mundie noted that the system could categorize them as employees based on their dress and change its behavior accordingly. For example, casually dressed people were assumed to be employees, he explained, and would be asked if they needed a shuttle. People appearing in a suit would be greeted in a different way.
Not only did the robotic receptionist extract context from what the subjects were wearing, but she could also extract information from a discussion between two subjects. When Laura asked if they needed a shuttle the men answered in the affirmative. When she asked which building they needed, a discussion between the men ensued as to whether or not it was building nine. The avatar listened, then inquired as to whether in fact they needed to go to building nine.
What’s amazing is the incredible amount of horsepower that was needed for the application. Just sitting idle at her virtual station, Laura burned up 40 percent of the compute power in an eight-core machine. But hardware isn’t the only obstacle to spatial computing, Mundie said. The software has become so complicated that it’s difficult to get such systems to work reliably and at the level of performance users will demand. Furthermore, the tools available for building such systems at a very large scale, with distributed components that work across the Internet, are inadequate. “Our tools were never designed to address this complexity,” Mundie said.