Implications of Watson on military operations
“Open the pod bay doors, HAL.”
— “2001: A Space Odyssey”
Released in 1968, the movie “2001: A Space Odyssey” depicts a super-intelligent computer, HAL, performing autonomous deep-space missions in collaboration with humans.
Fast-forward to 2011: The IBM computer Watson handily defeats two previous champions of the “Jeopardy!” game show. It appears that a profound change has occurred in the way humans and computers interact. On its website, IBM claims that the power of Watson has applications to the fields of medicine, finance and customer service.
This new computing capability has potential for military applicability, as well. If we can move past the cultural biases of evil super-intelligent computers as depicted in science-fiction movies, today’s military robotic concepts don’t go far enough to include future capabilities derived from sophisticated artificial intelligence (AI) converging with other technologies.
Current Army robotics strategy identifies only unmanned vehicles’ replacement of soldier tasks that are repetitive and/or dangerous in nature, based on technology of the near future. However, moon landings, PCs and lasers were examples of science fiction later to become reality. We are now at a crossroads where technological advances have moved beyond both concepts and policies regarding autonomous robots.
Given emerging technologies like a Watson, the intent of this article is to be a catalyst for discussion on the future implications of AI on operations across the war-fighting functions, and for the military to look again at policies on autonomous robots and its current robotics strategy.
Is the Army robotics strategy correct?
Because of rapid evolution in technology, the Army strategy toward robotics needs updating. While it was appropriate for technologies of the near term, computers like Watson were nonexistent at the time of concept development. Unquestionably, sophisticated AI converging with other potential technologies, such as non-silicon-based or nonbinary computing, will bring computers to another level, with significant ramifications for military operations.
The ad hoc nature of employment of unmanned ground vehicles (UGVs) and unmanned aerial vehicles (UAVs) in Iraq and Afghanistan is an example in the lag in concept development. Initially, commanders on the ground didn’t have a clear concept of how to employ unmanned systems. Now the Army is using more than 3,000 UGVs and has flown more than 400,000 sorties of UAVs in Iraq and Afghanistan. All are operated by a human operator in a safe location. This relationship could change once autonomous capabilities improve.
Army robotics strategy has its basis in multiple documents. The Congressional National Defense Authorization Act states that by 2010, one-third of the operational deep-strike aircraft of the armed forces will be unmanned, and by 2015, one-third of the operational ground combat vehicles of the armed forces will be unmanned. This is just over three years away, and yet there is little significant headway in attaining this goal.
Another key document is the Office of the Secretary of Defense Unmanned Systems Roadmap (2009-2034), in which four key missions are described: reconnaissance and surveillance; target identification and designation; counter-mine warfare; and chemical, biological, radiological, nuclear and explosive tactics.
The Army Capabilities Integration Center’s (ARCIC) Science and Technology Division and Tank-Automotive Research and Development Engineering Center jointly published a Robotics Strategy white paper in March 2009 in which they identified 32 soldier tasks to be replaced by robots in five categories: logistics, security, engineering, medical and maintenance. A feasibility analysis followed in which each task was rated as most-feasible, potentially feasible, and near-term infeasible, based on schedule, cost and complexity.
While the white paper does a good job of identifying potential robot tasks and implications along the DOTMLPF (doctrine, organization, training, materiel, leadership and education, personnel and facilities) domains, proposed concepts of robot usage do not fully account for the emerging technology of sophisticated AI. As an example, in the Robotics Strategy paper, the logistical task of “surface cargo transport and delivery of equipment and supplies using logistics convoys” is listed as a “near-term infeasible” robotic solution. The chief issues related to using robotics to enable vehicular surface movement include safety for nearby troops and civilians, battle command and awareness/tracking, integration with manned systems, and timely reaction to unexpected conditions (such as weather, obstacles and tactical conditions).
However, an autonomous robot could easily negotiate a dynamic route and protect itself throughout its entire journey. Even present-day technology allows for conduct of rudimentary convoy operations on both the surface and air on controlled routes.
Also from ARCIC, the Unmanned Systems Initial Capabilities document provides an initial step in identifying required capabilities across the war-fighting functions and recognizes the potentials for autonomy. However, the document highlights a weakness in robotics strategy in that currently deployed unmanned systems are remotely operated or semiautonomous, thus requiring a human to be an integral component for mission success. Recognizing autonomous robot capabilities is one thing. Developing, resourcing and then executing a comprehensive robotics strategy is another.
There are a number of potential military applications for AI. To identify opportunities, let’s briefly look at Watson. IBM describes it as “an application of advanced natural language processing, information retrieval, knowledge representation and reasoning,” and machine-learning technologies to the field of open-domain question answering that is built on IBM’s DeepQA technology for hypothesis generation, massive evidence gathering, analysis and scoring.”
What this means is that a computer that can think and learn from its mistakes and have a dynamic dialogue in a natural language with humans is a major step toward the required AI necessary for robots to operate autonomously while conducting complex military tasks. This is not only transformational, but within the realm of possibility. It is also a precursor for a robot of the future: not merely a tool, but a potential intellectual partner that understands war’s complexities. It could also track the multitude of data on the battlefield and then provide informed recommendations in free-form conversation for commanders to make decisions, much like the role of a human executive officer. However, this XO doesn’t need coffee to stay awake.
Undoubtedly, in its present form, Watson cannot perform tasks to military specifications. As the makers of the Internet could not fathom the transformational effects of their new invention, the potential of sophisticated AI applied to military usage is limited only by our imagination. The computer of today is like an abacus compared to the computer of tomorrow, and Army policies and concepts need to reflect the coming changes.
WATSON AND WAR FIGHTING
Let’s look at some practical AI applications through the lens of war-fighting functions. For purposes of illustration, let’s look at three functions and focus just on Army applications. As armed autonomy is prohibitively contentious, only nonlethal examples are provided:
Intelligence: A common problem is the ability to analyze the immense amount of data from multiple sources and turn it into timely, actionable intelligence. During counterterrorism operations, Watson could pick through relevant bits of information, recognize discernible patterns and then alert the human operator about suspicious activity. A capability like this would have helped authorities prior to 9/11.
Improvements in the understanding of language nuances, coupled with advances in facial recognition and understanding of human dimension traits that allow for the possibility to discern hostile intent, will be of immense value in identifying potential threats.
Shortages of translators in critical languages could also be a thing of the past because Watson will be conversant in hundreds of foreign languages and cultures.
Connected to a persistent-stare surveillance platform, Watson could proactively alert the operator if it knows what to look for, without ever getting tired.
Mission command: Since Watson represents increased autonomy, mission command could be greatly affected. Military units could operate in varying degrees of autonomy from higher headquarters. Whether they are full-robot or human-robot teams, the information warehousing, retrieval and analysis capabilities of mature AI, devoid of the need to communicate back to HQ for instructions, can assist in unit independence. This could have implications for locations of unit command posts or decentralization of decision-making down to lower levels.
In a human-robot team configuration, Watson as XO can provide data and make recommendations to the commander through management, retrieval and critical analysis of information. While disconnected or network degraded, its ability to warehouse large amounts of data such as terrain databases would allow Watson to accurately determine its location and provide route analysis, without a need to receive updates from a GPS satellite.
In a full-robot team configuration, autonomous robot teams can perform ISR missions and then report back only when needed, thus reducing the risk of detection or of taking up critical bandwidth.
Humans have a limited number of UAVs they could effectively control, but Watson would not have such a problem. As an example, executing a UAV swarm attack would be difficult for a single person. However, this problem could be resolved through an AI controller. If robots can control multiple robots more efficiently, is the next step for them to direct humans? Mission command would never be the same.
Protection: Reducing risk to the soldier is a key benefit of robots. Smart robots would be used to supplant humans doing dangerous or repetitive work. Automated surface convoys can reduce exposure of soldiers on vulnerable routes. Persistent-stare and increased alert awareness on surveillance platforms can significantly improve defensive capabilities.
Dangerous cleanup of hazardous materials is another role in which to replace humans. As displayed by the recent nuclear disaster in Japan, an autonomous robot could have ventured deep into the leaking reactor, diagnosed the problem, reported the status back, and then begun repairs without suffering radiation risk.
In force health protection, Watson could be a physician’s assistant and provide diagnosis based on symptoms cross-referenced with the extensive medical data in which even human doctors fail to maintain currency.
OBSTACLES
Fielding Watson still depends on solving difficult technical issues such as miniaturization, military hardening and consistent power. In its present form, Watson is an immobile, overgrown collection of servers in an air-conditioned building. To be tactically useful, Watson needs to be both mobile and tough enough to stand up in combat situations. Stable power supply, through improved batteries or even exploratory technologies such as regenerative bioconsumption, is also a possibility.
Watson still struggles with the understanding of a simple human trait like humor and is often stumped on word puns and double meanings. Watson won only 71 percent of its “Jeopardy” warm-up matches and even lost a match to a physicist following his successful Jeopardy challenge. The understanding of nuanced language is an area that needs improvement for Watson to be made useful for the military.
AUTONOMY AND CULTURE
Culture is an even bigger issue than technology. Perceived “Terminator-style” robots ranging the battlefield are still an anathema to U.S. culture. Policies regarding usage do not adequately address the advances in autonomy and the future capabilities that AI will allow.
Due to the lack of trust, having a human in the loop is still the default for armed robots, given the risk of killing innocent civilians. This is despite the fact that accidental civilian deaths have occurred with armed Predators operated by humans.
The medical community is also wary of autonomous robots treating humans. Even the new Battlefield Extraction-Assist Robot, or BEAR, is still manipulated by a human operator.
Another concern is the degree that soldiers are replaced by robots. The current intent is to supplant tasks, particularly dull and repetitive ones, to free soldiers for other missions. Hypothetically, what if robots become so good that soldiers are no longer needed? A military purist would argue that a human soldier can never be replaced by a robot. However, advancements in AI are proving that ever-increasingly complex tasks can be handled without human intervention. As an example, debates over a replacement helicopter for the Kiowa are shaped over the alternative solution of UASs. Where will all the helicopter pilots go? Nobody wants to be unemployed.
This argument is analogous to the post-World War I fight between the horse cavalry and the tank. In hindsight, the debate seems almost comical. We are once again at a crossroads of military development that technology has brought upon us.
In a time of budget constraint, investment in robotics will compete for other program dollars. In addition to the research and development and procurement costs, concept experimentation needs resourcing. The current Army budget provides $54 million per year for applied research and advanced technology development for unmanned vehicle technology. Additional funding is needed for future robotics programs of record.
KEEPING THE EDGE
While America holds a technological edge in information technology, this advantage may erode soon. Forty-three countries, including Russia and China, are developing military robot programs. In addition, commercial off-the-shelf solutions make it easier for nonstate actors to acquire robotics technology.
The Chinese pose the greatest threat toward the U.S. in the field of military robotics. They are not only technologically savvy, but are less constrained regarding moral or legal issues on the use of autonomous robots. Most alarmingly, the development of robotic asymmetrical capabilities in cyberspace and outer space leaves America particularly vulnerable.
The Chinese are known to establish front companies that actively steal robotics technology. They have already developed an unmanned aerial helicopter, the T-2, capable of carrying a small payload; it is an illegal copy of a Japanese version. Another UAV (ASN-229A) looks just like the Predator. Currently, the U.S. Army does not possess an effective counter to enemy UAS attacks.
To remain competitive, the U.S. military must develop a comprehensive robotics strategy. The necessary steps are:
Integrate the robotics community with concept developers. Many organizations within DoD are studying material solutions for robotics. However, there is no unified strategy or governance structure that moves away from the stove-piped approach and integrates concept development, requirements, and capabilities assessment. Since robotics has the potential of addressing every Army war-fighting challenge, every directorate should take an interest in forwarding robotics strategy.
Within the Army, future concept development happens at ARCIC at the Concept Development and Learning Directorate. ARCIC should establish a robotics task force that closely integrates with organizations in the robotics community such as the Defense Advanced Research Projects Agency, Robotic Systems Joint Project Office and the Joint Ground Robotics Enterprise, to name a few. The task force should consist of members from not only different directorates, but a cross-section of different generations and skill sets to provide alternate perspectives other than just those prevalent during past conflicts. The intent is to formalize required robotic capabilities and requirements through the Army war-fighter challenges and then develop the concepts for usage.
Conduct formal discussions on robotic usage. The DoD community needs to hold serious discussions on future robot strategy and the coming autonomy. The forum can take place in a senior-leader seminar, such as Unified Quest, or stand-alone panel discussions, dedicated to policy issues regarding robots, could be created. Discussions should focus on incorporating capabilities derived from sophisticated computing power and other emerging technologies. We need to analyze potential implications on the war-fighting functions, come to a consensus upon appropriate levels of autonomy, and then delve into the sticky moral and legal issues. The recently announced Autonomous Combat Systems Working Group run by the Office of the Secretary of Defense may address several of these issues.
Design a campaign of experimentation that encompasses forward-looking robotics concepts operating in a proliferated environment. There now is no doctrine on military robotics. To derive those concepts, a campaign of experimentation should incorporate scenarios that challenge assumptions and examine the utility of robots. It must be proven with analytical rigor that robots are a preferred solution to address capability gaps. The military should move beyond prototype experimentation by developing concepts and scenarios for experiments including wide usage of robots with sophisticated AI. This campaign of experimentation on unmanned systems can occur at the maneuver and aviation centers of excellence in sync with other agencies. There we can experiment on proper force mix/force design of robots and their appropriate autonomous capabilities.
Only digital soldiers die in experiments, making experimentation the perfect venue to test outside-the-box concepts regarding robotics. The British, French and Germans all had the tank during World War II, but only one of them was innovative enough to combine it with air and artillery assets, connected by the wireless radio, a tactic otherwise known as the Blitzkrieg.
Revise our current robotics road map. Technology in the field of robotics is advancing exponentially faster than our concepts. Revising our roadmap will have direct effect on requirements throughout the defense acquisition process and needs to be carefully analyzed. Training and Doctrine Command’s draft UGV campaign document recognizes possibilities for greater levels of autonomy for robots. Should policy decisions be made that allow for this greater autonomy, combined with the necessary technological advances, our future robotic strategy should reflect those changes.
Allocate resources for the effort. Well-funded programs of record for development are needed to execute a comprehensive robotic strategy. Without adequate funding or human resource allocation that encompasses robust research and development, experimentation, or testing and evaluation, executing a successful strategy is difficult.
While robots may never completely replace human soldiers, certain tasks are better suited for the robot. To keep up with the advancement in robotic capabilities, a coherent policy on autonomous usage is needed. Military robotics will improve only through the implementation of a comprehensive strategy that integrates concepts and capabilities development. The robot future is nearer than we think.
LT. COL. Anthony S. Cruz is chief of operations branch, Joint and Army Experimentation Division, of the Army Capabilities Integration Center, Training and Doctrine Command, at Fort Monroe, Va. The views expressed here are the author’s own and do not necessarily reflect those of the Army or Defense Department.
[…] programs and Flappy Birds. Army officer Anthony Cruz grasped the true potential of AI when he wrote about the need for a notional “robot general” that would take on higher-order cognitive functions in […]