Tech Innovation USA: From Resilient Networks To Self-Scheduling Devices
U.S. agencies are driving government IT initiatives that could soon make their way to businesses.
Beyond the brainpower of the workforce and community, there's "cognitive computing," the subject of a multiyear Defense Advanced Research Projects Agency initiative that's led to an automated office assistant and task management technology now being deployed by the military and commercialized by several startups.
Darpa's Personal Assistant that Learns (PAL) project started with the goal of tackling the challenges of machine learning. One result of that work is the Computer Agent that Learns and Organizes (CALO), an "office assistant" that takes input from scheduling, e-mail, and other software programs to create a knowledge base of a person's work habits. Over time, CALO learns user preferences by collecting that data and running pattern-finding algorithms against it.
"When you talk about meetings, the system learns, for example, that the user has a preference for the seventh floor and doesn't set up meetings between 11 and 1," says Robert Kohout, Darpa's program manager for PAL. Multiple CALO agents could potentially work together to hash out team or project schedules. The more data the agents collect, the better they get at determining preferences and accurately tweaking calendars.
Four CALO-derived projects are under way in the military, mostly of a sensitive but unclassified variety. One application prioritizes data for military intelligence purposes. The Army, meanwhile, is using what it learned from CALO to create a task management app that will learn how soldiers carry out routine tasks, such as generating event notifications or pulling information from a number of systems to create an incident report, and then automatically take on the most repetitive, simple steps without human input.
At the TSA, there might be a better way
PAL and other machine learning research could have implications for business computing. Another PAL project, Carnegie Mellon University's Reflective Agents with Distributed Adaptive Reasoning, focuses on crisis management. Researchers demonstrated an application that gathered information on how a conference organizer did his job and allowed an inexperienced substitute to manage an event based on the conference pro's preferences. "You could save millions of man-hours or billions of man-hours a year if you could do away with all the stuff that took low brain matter to do," Kohout says.
One of the reasons the PAL project has been able to maintain funding, Kohout says, is because Bill Gates once said that a breakthrough in machine learning could be worth 10 Microsofts. In fact, several companies have been launched based on findings from PAL, including "virtual personal assistant" provider Siri, whose iPhone app will be launched this summer. Initially, Siri will use technologies like speech recognition and natural language processing to help people do mobile Web search, but the company aims to move soon into building business apps.
IT Service Management Must EvolveThe idea of technology being delivered as a service appeals to the 409 IT pros responding to our Service-Oriented IT Survey. But cloud providers are competing for that work, and CIOs are being selective.