DARPA seeks deep-learning AI to cope with flood of information

Published 16 April 2009

The growing use of UAVs to loiter over enemy territory and send images and streaming videos back to HQ has created a glut of information; DARPA seeks a better, deeper, and more layered artificial intelligence to help the intelligence community cope with the avalanche of information coming in

For an intelligence officer, the only thing worse than having too little information is having too much of it. What with the increasing capabilities for collecting raw information — just think of the sheer quantity of images and video streams sent back by satellites and UAVs loitering the skies — there is a need to find a way to sift through this growing information hay stack in order to find the needles. DARPA — who else? —has a new plan to create powerful artificial intelligences. Lewis Page writes that the Deep Learning machines will be used to sift through petabytes of video from UAVs (we note the appeal of the animal kingdom for explanatory purposes: teachers in Junior High, explaining the facts of life to their students, talk about the birds and the bees; DARPA, explaining the differences between shallow and deep learning, talk about horses, cows, sheep, and goats).

According to DARPA, explaining the purpose behind the Deep Learning technology, the U.S. military and intelligence communities are drowning in surveillance and intelligence data. Hence the need for artificial intelligence to help them cope with the information flood:

A rapidly increasing volume of intelligence, surveillance, and reconnaissance (ISR) information is available to the Department of Defense (DOD) as a result of the increasing numbers, sophistication, and resolution of ISR resources and capabilities. The amount of video data produced annually by Unmanned Aerial Vehicles (UAVs) alone is in the petabyte range, and growing rapidly. Full exploitation of this information is a major challenge. Human observation and analysis of ISR assets is essential, but the training of humans is both expensive and time-consuming. Human performance also varies due to individuals’ capabilities and training, fatigue, boredom, and human attentional capacity.

One response to this situation is to employ machines …

There are already basic “shallow learning” AIs in use, including “Support Vector Machines (SVMs), two-layer Neural Networks (NNs), and Hidden Markov Models (HMMs).” These, however, are not much better than a human with poor “attentional capacity.” The trouble with the shallow learners is that they can learn only at a shallow level:

Shallow methods may be effective in creating simple internal representations … A classification task such as recognizing a horse in an image will use these simple representations in many different configurations to recognize horses in various poses, orientations and sizes. Such a task requires large amounts of labeled images of horses and non-horses. This means that if the task were to change to recognizing cows, one would have to start nearly from scratch with a new, large set of labeled data.

Page correctly points out that a specialized horse-spotter machine unable to recognize a cow is not much use for sorting the sheep from the goats. This is why DARPA wants “deeply layered” learning machines, able to identify horses, cows, sheep, and goats.

Deeply layered methods should create richer representations that may include furry, four-legged mammals at higher levels, resulting in a head start for learning cows and thereby requiring much less labeled data when compared to a shallow method. A Deep Learning system exposed to unlabeled natural images will automatically create high-level concepts of four-legged mammals on its own, even without labels.

U..S faces critical shortage of computer scientists

Published 13 January 2010

DARPA says the United States is facing a critical shortage of computer scientists; “While computers and internet connectivity become daily fixtures in the lives of Americans, we are steadily losing the engineering talent to [develop and maintain] these systems”

DARPA, the Pentagon’s research arm, issued a stark message the other day. According to the agency, the United States faces a crippling shortage of computer scientists in the near future, and only drastic action in the U.S. educational system can rectify this.

According to a DARPA solicitation:

The downward trend in college graduates with STEM [science, technology, engineering and maths] majors is particularly pronounced in Computer Science (CS). While computers and internet connectivity become daily fixtures in the lives of Americans, we are steadily losing the engineering talent to project these systems.

Lewis Page writes that far from gloomy employment prospects, DARPA reckons that IT types — especially computer scientists, particularly ones suited to military/DARPA style projects — are going to be increasingly in demand.

Our systems are becoming more complex, requiring more people with the software engineering talent to manage and maintain them. Finding the right people with increasingly specialized talent is becoming more difficult and will continue to add risk to a wide range of [military] systems that include software development.

Recent studies conducted by DARPA revealed that public perception is a critical issue. Study participants believed that the “dot-com bust” and “international outsourcing” have led to fewer computer science jobs. In fact, the opposite is true: the US Department of Labor lists “Computer & Software Engineers, Applications” as the fourth fastest growing occupation in the country in November 2007. Verbal reports from industry partners, as well as the presence of constant job openings, indicate industry is having difficulty finding software engineering talent to develop and maintain their software systems.

DARPA is in the habit of considering innovative, even radical new technology. A shortage of top-end, radical scientists, especially computing ones, is detrimental for the agency. Page notes that DARPA is not taking this lying down.

DARPA is interested in proposals with innovative new ideas to encourage students to major in CS-STEM and pursue careers as engineers and scientists. Increasing the number of graduates in Computer Science is a key goal, but the project will also be considered a success if the number of graduates in the broader STEM community is increased.

DARPA envisages this being done by reaching out to American kids as early as middle school and getting them interested in technology in the hope that they would finish college with a qualification useful for DARPA’s purposes. “In order to compel students to graduate with a CS-STEM related degree, it is important to maintain a positive, long term presence in a student’s education,” notes the agency.

BAE Systems completes first flight test of persistent surveillance

Published 10 February 2010

The ARGUS-IS offer a new real-time persistent surveillance capability for U.S. combat forces to detect, locate, track, and monitor events on battlefields and in urban areas -- providing significantly greater video coverage over current airborne capabilities

BAE Systems has completed the initial flight test of a new real-time persistent surveillance capability for U.S. combat forces to detect, locate, track, and monitor events on battlefields and in urban areas — providing significantly greater video coverage over current airborne capabilities.

The first flight tests of the Autonomous Real-Time Ground Ubiquitous Surveillance Imaging System, or ARGUS-IS, occurred aboard a U.S. Army Black Hawk helicopter. The tests demonstrated the system’s multiple video windows for persistent area surveillance and tracking capabilities for vehicles and dismounted soldiers.

The airborne processing system can simultaneously and continuously detect and track the presence and motion of thousands of small or large targets over an area covering tens of square miles. BAE Systems designed and produced the system’s sensor and processor.

“ARGUS-IS will significantly advance the Army’s capability to protect its troops through improved search and surveillance capabilities,” said Dr. John Antoniades, ARGUS program manager and director of remote sensing technology for BAE Systems.

BAE Systems equipment aboard ARGUS-IS consists of a high-resolution, extreme wide-area, real-time video sensor; an on-board processing system; and ground processing for interactive multi-target designation, tracking, and exploitation. “The ARGUS-IS system overcomes the fundamental limitations of current airborne surveillance systems,” said Dr. Steven Wein, director of optical sensor systems at BAE Systems. “Very high-resolution imaging systems required for vehicle and dismount tracking typically have a ‘soda-straw’ view that is too small for persistent coverage. Existing wide-area systems have either inadequate resolution or require multiple passes or revisits to get updates.”

The Defense Advanced Research Projects Agency (DARPA )and the U.S. Air Force Research Laboratory awarded BAE Systems an initial $18.5 million contract to lead the ARGUS-IS effort in late 2007. The system is targeted for use in Department of Defense unmanned and manned surveillance platforms.

BAE Systems has approximately 105,000 employees worldwide, and its sales exceeded $34.4 billion in 2008.

LGS on Lockheed Martin team for $31 million DARPA cyber assurance contract

Published 18 February 2010

LGS selected by Lockheed Martin as a subcontractor for a 31 milllion dollar DARPA-funded contract to develop cyber procedures which will provide military untis with dynamic bandwidth allocation

LGS Innovations, a subsidiary of Alcatel-Lucent, announced that it was named a subcontractor on the Lockheed Martin team that was recently awarded a contract from the Defense Advanced Research Projects Agency (DARPA). The $31 million Military Network Protocol contract calls for Lockheed Martin to develop new cyber assurance procedures that will improve security, and provide dynamic bandwidth allocation and policy-based priorities at the individual and unit levels.

The LGS Internet Research Department will work as a part of the Lockheed Martin’s Infrastructure team whose goal is to define, design, develop, and test the common elements used in the DARPA solution.

“Our country’s increasing dependence on information technology has made cyber security one of the most critical challenges facing the U.S. military today,” said Dom Imbesi, director of the Internet Research Department at LGS. “LGS is pleased to be a part of Lockheed Martin’s strong team working to address and solve these challenges.”

In developing this new protocol for military networks, Lockheed Martin’s team will develop router technologies that will include strong authentication and self-configuration capabilities to improve security, reduce the need for trained network personnel and lower overall life cycle costs for network management.

The Internet Research Department is the R&D division of LGS dedicated to the fields of Computer Network Operations (CNO) and Information Assurance (IA). The department has a history of achievement as formerly being a part of the renowned Bell Labs and can utilize the technical expertise and resources of Bell Labs to provide cutting-edge ideas and technology to solve today’s toughest cyber security problems. The LGS capabilities run the full spectrum of CNO activities: Network Reconnaissance, Host Fingerprinting, Attack Planning, Payload Delivery, and Battle Damage Assessment. The department also performs leading-edge Information Assurance research and development with a specific focus on mission assurance, external exposure assessment and predictive analysis.

DARPA looking for military iPhone and Android apps

Published 3 March 2010

Pentagon's research arm is looking for apps to be written for the iPhone or for handsets running Google's Android OS -- "with potential relevance to the military specifically and the national security community more generally"

DARPA has announced that it would like some apps written for the iPhone or for handsets running Google’s Android OS — “with potential relevance to the military specifically and the national security community more generally.”

The Pentagon researchers note that:

In today’s military, handheld systems are characterized by a tight integration of specialized hardware with a narrowly focused software suite. Most of the handheld devices are heavily optimized for a particular task and are ill-suited for general-purpose use. A soldier’s radio, for example, has very limited data capability and essentially no multimedia capability. Current language translation devices support neither messaging nor collaboration of any form …

A transformation in technical approaches and business processes is called for.

Lewis Page writes that it will not be a transformation powered by Windows Mobile. DARPA specifies that “initial interest will focus on apps developed on the iPhone or Android platforms.” The idea is to find apps which will be helpful “especially among the end-users at lower levels in the military echelon.”

There are already apps for the iPhone which can make ballistic calculations for a sniper. Other existing software which would help a soldier could include various kinds of navigation kit, user interfaces for remote drones — several of which are already on offer - and so on.

Page notes that some military hardware, too, has already taken on many of the aspects of a smartphone — for instance, the Land Warrior wearable comm/puter rig. DARPA’s preferences notwithstanding, at least one maker has produced a covert version of military belt-computer software to run on a Windows smartphone.

DARPA, unusually, would seem to be very much with — or even a bit behind — the times on this one. The agency normally prefers to be well into the future.

Full instructions for developers to submit whitepapers to DARPA can be found here.

The last frontier: DARPA wants to make the Earth's crust transparent

Published 10 March 2010

Seeing through the Earth's would allow the development of tools to protect civilian populations from the ravages of natural disasters; these same tools could be used for military purposes against enemies -- detecting, targeting, and destroying hard and buried underground facility (UGF) targets

DARPA, that intellectually restless agency where, according to Lewis Page, they “believe it is better to invent a head-mounted multispectral imaging device than curse the darkness,” is pushing the envelope again. The agency has been interested in mastering — some would say lording over — the nature for years now. The agency talked about planet hacking and influencing enemy climate (aka “weather war”), and the agency still wants to harness the power of lightning.

Katie Drummond writes that this year the agency has an agenda which is no less ambitious. As part of its budget for the upcoming fiscal year, DARPA is launching the Transparent Earth project. The agency will invest $4 million into the creation of real-time, 3-D maps that display “the physical, chemical and dynamic properties of the earth down to 5 kilometer depth.”

Drummond notes that at first, the idea does not sound all that impressive. The earth is more than 3,500 miles deep, from crust to core, so DARPA’s plan would not do much more than scratch at the surface. Geologists and geophysicists, however, still know very little about the day-to-day goings-on underground, even at a depth as shallow as 5 km. The deepest drilling of the planet was a Soviet hole on the Kola Peninsula, which took nineteen years and made it around 7.5 miles into the crust, and even NASA still uses land-based GPS signals to predict volcanic eruptions.

Now, rather than a massive drilling project, DARPA wants to harness innovations in sensor technology to develop a constantly updating model of planetary activity. The project will use sensors to detect “natural indicators of subsurface activity,” and then take advantage of mathematical algorithms designed to estimate various natural earthly phenomena, including geophysical turbulence and shifting tectonic plates.

Drummond writes that algorithms are already used in planetary mapping and predictive science, but adding high-tech sensors would provide a constant stream of new data. “This kind of accuracy could have serious planetary implications: Changes in the earth’s crust can explain and predict volcanic eruptions, earthquakes and even the formation of mountain ranges,” she writes.

After they successfully combine sensors and mathematics, DARPA’s ultimate goal would put even NASA’s satellite footage to shame: A global three-dimensional picture of the earth’s subsurface with variable spatial, temporal, and information resolution, allowing changes at local scales to propagate through both physical models and proximity rules to update the global picture.

So maybe DARPA wants to protect civilian populations from the ravages of natural disasters, Drummond writes. These same tools, however, could be used for military purposes against enemies, Drummond quotes one unnamed geoscientist to say. “All of my ’science is good!’ tree-hugging comments aside, what this program is probably really about is detecting, targeting, and destroying hard and buried underground facility (UGF) targets,” the geoscientist said.

The last frontier: DARPA wants to make the Earth's crust transparent

Published 10 March 2010

“Whatever DARPA’s intention, they want their transparent earth sooner rather than later: The agency anticipates that the new 3-D models will be available to the Army, Air Force, special operations and intelligence agencies by 2015,’ Drummond concludes.