DoD BEGINS RESEARCH INITIATIVE TO DEVELOP ‘COGNITIVE’ NETWORKS
Dept. of Defense (DoD) unveiled details of how it wanted to develop advanced information-management technology with help of private sector researchers. Defense officials in recent months have expressed need for unique approaches in how to process and intelligently disseminate data over DoD’s evolving Global Information Grid. Defense Advanced Research Projects Agency (DARPA) last week issued formal call for proposals to develop “cognitive” devices and systems, or machines that “know what they are doing” and can streamline delivery of voice, video and data communications to warfighters and policymakers. DARPA, whose predecessor agency invented Internet, acknowledged there were major research challenges to overcome before cognitive IT processing became reality. However, once such processing were attained, it would “provide substantial new opportunities in all sectors, encompassing critical defense, industrial and commercial applications,” DARPA said.
Sign up for a free preview to unlock the rest of this article
Communications Daily is required reading for senior executives at top telecom corporations, law firms, lobbying organizations, associations and government agencies (including the FCC). Join them today!
Initiative reflects concerns expressed by David Kelley, Lockheed Martin Mission Systems vp-Information Operations. Speaking last week at Armed Forces Communications & Electronics Assn. (AFCEA) convention in Washington, Kelley said govt. and industry must expand their vision of future networks beyond advances in technology standardization and equipment interoperability (CD June 12 p5). Development of machine-to-machine communications processes, which intelligently collect and disseminate data to warfighters, must become priority, he said. Major Gen. Steven Boutelle (USA), dir.-Information Operations, Networks & Space, said at AFCEA event that Pentagon would look to private sector for help in developing such advanced information-management technology.
Ability of troops to communicate seamlessly via land, sea and satellite systems around globe, which DoD refers to as network-centric warfare, is at core of current Defense- communications modernization efforts. However, DoD is aware that it needs to develop networks that handle vast amounts of information without overwhelming recipients of those data. DARPA’s Information Technology Processing Office (IPTO) said: “The focus of our new direction in IPTO is on ways to make systems not simply faster and smaller, but smarter… If we could truly build systems capable of practical reasoning, learning and self-awareness, we would expect that such systems would easier to extend and maintain… Intelligent user interfaces could adapt to their users, rather than the other way around.”
Despite “tremendous gains” in computational power, specifically by doubling number of transistors on chip roughly every 18 months, “we are increasingly being defeated by that complexity… The raw power that has so seductively invited us to build systems of unprecedented size and complexity has led to systems… that regularly fail in practice and are increasingly vulnerable to attack,” DARPA said.
Technological areas that DARPA is interested in developing include “computational perception,” in which sensors and other interfaces accept and filter data from numerous sources, then discard “noise and clutter” before routing data. DARPA said it also was interested in new processes that “use a cognitive system’s knowledge and experience to influence its perceptual capabilities.” It said revolutionary advances in that area would enhance ability of network “to detect important objects, events and situations in its environment.”
Development of advanced interfaces among cognitive systems, legacy systems and humans also is priority, DARPA said: “Of special interest are capabilities that allow cognitive systems to be instructed, guided and persuaded using natural human oriented communications,” such as natural language, pictures, gestures. Systems with multiple components “that can achieve goals in a coordinated way” also are sought by DARPA. These “cognitive teams” as it envisioned would be composed of “artificial cognitive agents, humans and noncognitive components,” and would require new technology “that facilitates the successful performance” of the teams, it said.
DARPA wants to establish underlying scientific and mathematical foundations for cognitive computing where none exist or for where they exist but currently are inadequate. It said it was interested in “technologies for biologically inspired computing” that would integrate elements of “neuroscience and neural systems, semiconductor implementations, and realizable architectures.” Abstracts are due April 4, 2003, research proposals June 6, 2003 -- 703-741-7804.