DoD collects loads of data from satellites, drones and Internet-of-things devices. But it needs help making sense of the intelligence and analyzing it quickly enough so it can be used in combat operations. But, the sheer volume of video content produced makes identifying, assembling and delivering actionable intelligence — from multiple sources and across thousands of hours of footage — a habitually long, laborious process.
The DoD is now developing an AI-driven algorithm to work in conjunction with its drone footage to spot, tag and bookmark potential threat targets. Military analysts are using Google-developed AI algorithms to mine live video feeds from drones. With machine learning techniques, software is taught to find particular objects or individuals at speeds that would be impossible for any human analyst. This AI technology can differentiate between people, objects and buildings, much like Google’s driverless cars. Undersecretary of Defense for Intelligence Joseph Kernan said Project Maven only started a year ago and so far has been “extraordinarily” useful in overseas operations.
Project Maven, also known as the Algorithmic Warfare Cross-Function Team, was launched in April 2017 by then-Deputy Defense Secretary Bob Work to accelerate the department’s integration of big data, artificial intelligence and machine learning into DoD programs. Project Maven is a Pentagon project using machine learning to sort through masses of intelligence, surveillance and reconnaissance data – unmanned systems video, paper, computer hard drives, thumb drives and more – collected by the department and intelligence agencies for operational use across the services. “As numerous studies have made clear, the department of defense must integrate artificial intelligence and machine learning more effectively across operations to maintain advantages over increasingly capable adversaries and competitors,” Work wrote.
Among its objectives, the project aims to develop and integrate computer-vision algorithms needed to help military and civilian analysts encumbered by the sheer volume of full-motion video data that DoD collects every day in support of counterinsurgency and counterterrorism operations. Project Maven focuses on computer vision — an aspect of machine learning and deep learning — that autonomously extracts objects of interest from moving or still imagery, Cukor said. Biologically inspired neural networks are used in this process, and deep learning is defined as applying such neural networks to learning tasks.
Drew Cukor, chief of the DoD’s Algorithmic Warfare Cross-Function Team, said in July: “People and computers will work symbiotically to increase the ability of weapon systems to detect objects. Eventually we hope that one analyst will be able to do twice as much work, potentially three times as much, as they’re doing now. That’s our goal.” It has reportedly already been put into use against Islamic State. He said the immediate focus is 38 classes of objects that represent the kinds of things the department needs to detect, especially in the fight against the Islamic State of Iraq and Syria.
SOCOM Commander Army Gen. Richard Clarke said that, in recent conversations he has had with commanders in Afghanistan, between 2001 and 2018 “70 percent of their time was putting bomb on target.” But now, through better use of AI for speedy and precise targeting and a greater understanding of the effectiveness of information operations, they have reversed that ratio. In a combat situation, Clark said, “you’re not trying to kill every bad guy out there” but rather are targeting a leader or a group of leaders. AI has already gained a strong foothold in logistics and maintenance in Pentagon thinking and is now making its way to commanders. “Maven has made some inroads [because] it is actively giving them courses of actions” and even parallel courses of actions to take simultaneously to further confuse an enemy.
Mohsen Fakhrizadeh, who founded Iran’s nuclear program in the 2000s, had a security detail of 11 guards while traveling with his wife on Nov. 27 in a car on a highway outside Tehran when an automatic machine gun outfitted with AI and an advanced camera zoomed in on his face and fired 13 times, an Islamic Revolutionary Guards Corps deputy commander told local media Sunday. The assassination of Iran’s top nuclear scientist in Nov 2019 was carried out remotely with artificial intelligence and a machine gun equipped with a “satellite-controlled smart system”, Iranian news agencies quoted a senior Iranian commander as saying.
Ali Fadavi, the deputy commander of Iran’s Revolutionary Guards Corps, told Iranian news agencies that Mohsen Fakhrizadeh was driving when a weapon opened fire on his car on a highway near Tehran. The weapon “zoomed in on Fakhrizadeh” using an “advanced camera”, Fadavi said. “No terrorists were present on the ground.” Fadavi said the gun used to kill Fakhrizadeh had been placed on a pickup truck and controlled by a satellite, and had fired 13 shots. “During the operation artificial intelligence and face recognition were used,” Fadavi said. “His wife, sitting 25cm away from him in the same car, was not injured.”
This incident points to many advanced technologies like Maven algorithms that might have been applied on images captured by advanced camera, precision fire of automatic machine gun and remote control operation through satellite.
When Maven was introduced it did not perform according to senior military leaders, but as the machine learning programs got better with more training data it became to get trusted by military analysts and adopted with many arm of DOD. With that being said, Schultz added that the Army’s 82nd Airborne Division and its XVIII Corps “want to be the first” division and corps to apply AI and machine learning across the board, from planning to operations to sustainment. He said the commanders had a special operations background before moving into their new positions, and that experience made them more familiar with what these technologies could do for them.
“The way Maven works is, it’s like your three-year-old plays with his iPad,” head of Air Force Air Combat Command Holmes told reporters. The child looks at one image after another and says, “It’s car, car, car, not car,” the general said. “You work through it, have to teach it, and it learns and it’s learning — but it hasn’t learned to the point where you still don’t have to go back and have mom or dad look over the shoulder of the three-year-old and say, ‘yeah, that really is a car,’ or ‘yeah, that really is green.’”
Maven algorithms are training themselves and getting better and better, which is exactly what machine learning is supposed to do: Feed the system more and more data and it gets better at what it’s supposed to do. The Air Force is already relying on AI to help with predictive maintenance and other tasks, Holmes noted, but those efforts have already been well tested by commercial airlines.
Another reason Project Maven is “disruptive” is that it shows that analysts are beginning to trust new sources of intelligence and nontraditional methods, Manzo said. “What’s encouraging is that the outputs of these systems are being trusted by the users,” he said. “A machine comes up with an answer and the human gives the thumbs up or down,” he said. “If DoD is trusting this, it’s a tremendous step.” Even though a human is supervising, the focus doesn’t have to be on “making sure the machine is doing the things I asked the machine to do.”
Army AI Task Force Project as Maven Offshoot
The U.S. Army is developing data surveillance and analysis technology in Pittsburgh, though an offshoot of the controversial Department of Defense initiative Project Maven within Carnegie Mellon University’s Army AI Task Force. CMU’s involvement comes after the project made headlines last year when Google employees signed a petition en masse to cease work on the initiative. The project, the second undertaken by the Task Force, is seeking to develop an algorithm to analyze drone, overhead, and ground data to identify targets and objects of interest.
The military does so much reconnaissance that human analysts simply can’t keep up. Their image and sensor networks generate tons of data daily, which the Department of Defense says necessitates the use of artificial intelligence to do effective analysis. Project Maven is the DoD’s national initiative to solve this problem, which the Task Force’s project is tied to.
Colonel Doug Matty, deputy director of the Army AI Task Force, says upon completion the project will allow them to “generate effects without kinetic mechanisms.” In other words, the Army hopes this system of analysis will be able to surveil from miles above, without having boots on the ground.
“You often hear people talking about operating at the speed of cyber, well, we see that artificial intelligence will allow us to operate at those kinds of speeds,” Matty said. “It’s no longer just, I’d say, the ‘X, Y and Z’ of ‘shoot, move, communicate,’ kinds of things. We’re now leveraging not just the physical areas, whether its land, air, sea, space, but also cyber.” The computer program will assist commanders in making decisions, but will not make them itself.
According to Army Times, a key guide for the current Maven affiliated project was a directive from former Secretary of the Army Mark T. Esper and Army Chief of Staff Mark A. Milley to develop long-range precision fires, Matty explained. He says with this Maven project, “what we’re able to do is leverage this kind of advanced technology to support, you know, those kind of mission threads.”
Air Force moving Project Maven into Advanced Battle Management System portfolio
The program office for Project Maven — the Air Force’s first major foray into using artificial intelligence to scan drone footage — is being transitioned into the Advanced Battle Management System (ABMS) portfolio as the service continues positioning more of its traditional backend IT capabilities to support broader warfighting functions.
The Air Force is working to fold Project Maven into its ABMS tech stack to use the program’s AI capabilities to analyze and link data from the vast array of sensors used in battle. ABMS is the technical backbone of the joint force’s concept of Join All-Domain Command and Control (JADC2) — a network-of-networks that aims to link “every sensor to every shooter” across air, land, sea, space and cyber.
The Air Force is transitioning Project Maven to align AI-efforts and bring new capabilities to machine-to-machine data sharing, Will Roper, the Air Force’s assistant secretary for acquisition, technology and logistics, said Friday. “We are bringing Maven capabilities into the developing tech stack for ABMS.”
ABMS is similar to Maven in that it aims to use AI to improve the coordination of so-called kill chains and remove humans from tedious tasks. The in-development platform, however, focuses more broadly on networking and communications systems across domains to support military operations. The plan is that AI-enabled systems will process raw data collected from battle, turn it into actionable insights and push it to the people in the chain of command who need it.
The Air Force is also funneling many of its other key IT platforms into ABMS development. For instance, Cloud One and Platform One are both now supporting ABMS cloud and enterprise software development needs. Platform One was recently designated as the Pentagon’s first enterprisewide DevSecOps platform with which coders across the services can develop products to meet mission needs. The platform is used, along with cloud services from Cloud One, in other programs across the Air Force.
The department is now full-steam-ahead on using both platforms for the most advanced warfighting system development, Roper said. “There is no distinction between development systems and warfighting systems anymore in IT,” he said. “ABMS and Maven are going to start blurring that line in September.”
Maven was created by Google
Maven was created by Google — which later abandoned the project — and other companies and was meant to be the leading edge of the Pentagon’s efforts to make AI useful. It got tools to the troops within six months in 2017. Intelligence officials are enthusiastic about Maven in its current form, because it is automating the mundane process of looking at the massive amounts of video from reconnaissance aircraft and other assets and trying to make sense of it in a predictable and rapid fashion
Google’s TensorFlow AI systems are being used by the US Department of Defense’s (DoD) Project Maven, which was established to use machine learning and artificial intelligence to analyse the vast amount of footage shot by US drones. The initial intention is to have AI analyse the video, detect objects of interest and flag them for a human analyst to review.
According to a new report from The Intercept, Google hired gig economy workers to help build out a controversial artificial intelligence program that the company had paired with the Pentagon to build. The workers were hired through a crowdsourcing gig company outfit called Figure Eight, which pays as little at $1 an hour for people to perform short, seemingly mindless tasks. Whether the individuals were identifying objects in CAPTCHA-like images, or other simple tasks, the workers were helping to train Google’s AI that was created as part of a Defense Department initiative known as Project Maven.
By employing these crowdsourced microworkers, Google was able to use them to teach the algorithms it was running how to distinguish between human targets and surrounding objects. According to The Intercept, these workers had no idea who their work was benefitting or what they were building.
A Google spokesperson said: “This specific project is a pilot with the Department of Defense, to provide open source TensorFlow APIs that can assist in object recognition on unclassified data. The technology flags images for human review, and is for non-offensive uses only.”
Figure Eight, which was previously known as Crowdflower, is one of the largest platforms that employs microworkers. On its website, Figure Eight says its platform “combines human intelligence at scale with cutting-edge models to create the highest quality training data for your machine learning (ML) projects.” By partnering with these microworker outfits, Google could quickly and cheaply build out its AI. “You upload your data to our platform and we provide the annotations, judgments, and labels you need to create accurate ground truth for your models,” the website reads.
The AWCFT has already delivered the first algorithms to warfighting systems by the end of 2017. The team is already developing proposals to take on the next set of challenging intelligence projects. In Phase II, Project Maven will expand its scope, turning the enormous volume of data available to DoD into actionable intelligence and decision-quality insights at speed.
Defense procurement chief Ellen Lord said the Pentagon will start bringing together AI projects that already exist but do not necessarily share information or resources. “We have talked about taking over 50 programs and loosely associating those,” Lord told reporters. “We have many silos of excellence.” Undersecretary of Defense for Research and Engineering Michael Griffin will oversee a new AI office that will bring in “elements of the intelligence community,” he said. But many details remain to be worked out.
Google’s involvement with Project Maven stirred a controversy inside the tech giant. The company had signed a contract with the Defense Department to develop artificial intelligence that could interpret video images in order to improve drone targeting. But after the contract’s disclosure sparked an internal rebellion among employees, Google allowed its contract to expire. The deal is set to end in March 2019.
Defense tech startup founded by trump’s most prominent Silicon Valley supporters wins secretive military AI contract
A startup founded by Palmer Luckey, the 26-year-old entrepreneur best known for having founded the virtual reality firm Oculus Rift is among the latest tech companies to win a contract with the Pentagon as part of Project Maven, the secretive initiative to rapidly leverage artificial intelligence technology from the private sector for military purposes.
Anduril Industries is developing virtual reality technology using Lattice, a product the firm offers that uses ground- and autonomous helicopter drone-based sensors to provide a three-dimensional view of terrain. The technology is designed to provide a virtual view of the front lines to soldiers, including the ability to identify potential targets and direct unmanned military vehicles into combat. The first phase of the research has been completed, according to the documents reviewed by The Intercept, with initial plans to deploy virtual reality battlefield-management systems for the war in Afghanistan.
“What we’re working on is taking data from lots of different sensors, putting it into an AI-powered sensor fusion platform so that you can build a perfect 3D model of everything that’s going on in a large area,” Luckey said. “Then we take that data and run predictive analytics on it, and tag everything with metadata, find what’s relevant, then push it to people who are out in the field.”
“Practically speaking, in the future, I think soldiers are going to be superheroes who have the power of perfect omniscience over their area of operations, where they know where every enemy is, every friend is, every asset is,” he said. Luckey said he thinks it is “unlikely” that soldiers of the future will directly carry weapons in the field; instead, they would remotely operate machines and weapons from far away.
Anduril previously garnered attention for its efforts to help U.S. Customs and Border Protection create a “virtual wall” at the U.S.-Mexico border. The initial 10-week demonstration used Anduril’s Lattice technology to monitor a stretch of land along the Rio Grande Valley. The system reportedly helped the government identify and apprehend 55 unauthorized individuals crossing the border.
The company has also publicly acknowledged work to develop perimeter defense monitoring around two U.S. Marine bases.
Anduril’s pitch deck, the presentation it provided to solicit investors, imagines a future of warfighting by means that might look like science fiction to the average observer. The company is pushing battlefield management technology capable of utilizing long-range bombers and swarms of military attack drones. The firm has reportedly rented a warehouse in Oakland, California, to develop at least one remote-control tank, designed for fighting California wildfires.
In an era in which the Department of Defense is criticized for delivering solutions too slow, one effort on the cutting edge of technology is proving the opposite.
Aside from just being a pathfinder project to solve a critical need of more quickly processing intelligence using machine learning, Project Maven is “proving out how we go fast and how we deliver to the field,” Kari Bingen, deputy under secretary of defense for intelligence, said Sept. 5 at the Intelligence and National Security Summit hosted by INSA and AFCEA.
“We’re not here talking about come see me in five years and we’re finally deliver something to the user downrange,” Bingen said when asked what excites her about the effort. “This is six months from authority to proceed to delivering a capability in theater.”
Touting the same glee in Maven’s rapidity, Bingen’s boss Joseph Kernan explained earlier this year how Maven was under contract within two months and it actually delivered capability within six months.
All agreed at Hudson Institute panel on handling big data in military operations during online forum that, even with the adaptation of AI and machine learning to military operations, forward presence would remain important. How many forces, however, could be reduced when forces like Marines have more tools in their hands to succeed. Clark added, “we’re definitely seeing that synergy” in the Marine Corps on applying AI to operations. “We can prevail” against the Chinese and the Russians in the new arms race, Schultz said. “We just need to be able to harness [artificial intelligence].”
References and Resources also include: