News

2021-03-01

Naval Embrace of AI Needs to Tighten

Navies around the globe are increasingly realising that ignoring the limitless benefits offered by Artificial Intelligence (AI) in naval operations could land them in troubled waters.

While the navies play a prodigious role in the protection of nations, ensuring safety of the seas/ maritime routes and involving in humanitarian interventions, the challenges they face are complex and multiplying by the day.

Hardening of threats, acceleration of technological developments, new economic and environmental imbalances and the sheer size of the ocean environment are prompting navies to reassess their options.

The message out there is clear: The groundwork for operations with AI can—and must—be laid today, or the naval forces of the nation will be left unprepared for future missions.

There is no time to pause. Current business practices can be implemented expeditiously within the navy to reduce costs, increase efficiencies, generate new capabilities, and reduce manpower requirements in noncombat roles, which would help increase the number of sailors and Marines available for deployments and operations.

The practical applications of AI involve intricate tasks that augment human involvement and increase humankind’s own abilities and productivity.

AI systems are already helping organisations respond to humanitarian disasters and emergencies.

Evolving with each passing day, the definition of artificial intelligence itself has changed many times since the first conference on AI at Dartmouth College in 1956, at which researchers joined together to theorise about the combination of robotics, neural networks, and programming.

At present, AI is a self-teaching machine. Rather than a programme with set inputs and outputs, AI teaches itself and changes as its environment changes.

4 Layers of AI System

An AI system has four layers, which interact with each other to mimic human intelligence. AI itself is considered the topmost layer, which absorbs, stores, and processes information to make decisions.

One layer below, the AI relies on machine learning, which allows it to “learn and act without the need for human input.”

The third level is deep learning, which contains the AI’s ability to process images, speech, and language.

Finally, the bedrock of the AI system is the neural network, which processes data. The most significant opportunities for new research exist in this neural-network layer.

It’s not that AI can supercede human intelligence in the immediate instant. While the human brain has over one hundred billion neurons, the most advanced AI available today only has about one billion neuron equivalents.

Artificial Intelligence itself is divided into two categories.

General AI attempts to mimic the human brain in autonomous thought, while narrow AI is the creation of smart computers to deal with complex problems.

Though General AI does not exist as yet, major progress in the field of narrow AI provides huge opportunity for the eventual creation of general AI.

In fact, AI, as many understand and use today, is narrow AI.

For example, narrow AI is used on most commercial passenger planes. On Boeing 777s, pilots only spend about seven minutes out of every flight manually flying the plane, while Airbus pilots manually fly about three and a half minutes of every flight.

IBM’s Deep Blue and Watson projects are both advanced versions of narrow AI that have received much attention. Amazon’s Alexa and Apple’s Siri are both examples of narrow AI beginning to make significant impact on people’s lives.

Advanced progress in narrow AI can be used with human-in-the-loop (HITL) systems for “expanded human potential.” While AI systems have beaten world-class chess players on numerous occasions, the greatest success is achieved when an AI is paired with a human. In an HITL system, human decisions and operations are advanced through integration with AI, such as in flight simulation trainers. An HITL system requires a human user working with the AI and making decisions on the basis of AI recommendations. The system empowers human interaction with AI. Importantly, the platform can be designed to defer decision-making to human operators.

Unmanned aerial vehicles and other intelligence, surveillance, and reconnaissance (ISR) platforms operate with these designs. Both HITL and human-on-the-loop systems—in which a human becomes involved to override the system when needed —are certain to play a vital role in future military applications of AI.

Scope for Naval Application

The best examples of AI that can apply to the Navy are those that people are already using effectively. Simple AI programmes are all around us. Google Maps uses AI to programme the most-efficient routes for drivers. Chatbots such as Siri, Alexa, and Microsoft’s Cortana are AI applications that have advanced considerably and continue to learn and refine their output to support the personalised needs of their users better.

AI can even act as an “intelligent salesman,” providing personalised sales recommendations to customers—also known as smart advertising. Organisations such as Uber Technologies use dynamic pricing—accurately pricing a commodity or service between supply and demand.

AI engines can create fake videos with realistic images and sounds. Researchers at the University of Washington were able to create a realistic but fabricated video of former president Barack Obama giving a speech using AI.

Navy and Marine Corps could use it to gain insight into an individual’s actions and thoughts relevant to warfare.

DoD’s Unmanned Aircraft

The integration of greater AI autonomy reduces casualty rates of armed forces’ personnel. Such systems can adopt riskier tactics, target with greater accuracy and operate with greater endurance, range, and speed even while retaining a superior level of flexibility and mobility

It will be interesting to note that the U.S. Department of Defence’s (DoD) 11,000 unmanned aircraft currently make up 40 per cent of the total number of U.S. military aircraft.

There are six mission parameters determining the applicability of AI. These are speed of decision making, volume of data, quality of data links, complexity of the action, danger of the mission, and required endurance.

AI can be a critical component in missions with high or complex levels of these parametres, such as cyber operations; missile defence; data analysis; ISR data integration; contested communication or operations; unmanned vehicle operations, including unmanned undersea operations; air operations centre activities; multi-mission operations; and chemical, biological, radiological, and nuclear attack cleanup.

Near-Term AI Applications

The near-term applications of AI can make the Navy more responsive and deadly. Since narrow AI is simply the composition of “machine-learning solutions that target a specific task,” the technology can be applied to a range of functions, especially in noncombat processes.

These near-term AI applications can cut costs and lay the groundwork for full-scale adoption of complex AI systems in the next decade.

Currently functioning AI systems can be used to support administration, personal productivity, planning, logistics, crisis response, training, intelligence, force protection, and force structure.

This section will examine these possibilities by applying current private industry practices to naval functions.

Administration

Administration is a primary support function within the Navy that can be revolutionised through AI. Numerous companies already use AI to assist with information-input management, which involves processing “incoming mail, e-mails, invoices, spreadsheets, presentations, PDFs, and other documents.” AI can help with the preprocessing of information (i.e., who needs this information and how does it reach them) as well as the maintenance and categorisation.

Insurance providers use AI to identify topics from e-mails and letters and route them to the correct internal departments. Logistics companies are using AI to assist with internal functions such as accounting and human resources.

AI can work with robotic, rules-based processes, such as filling in forms and accessing data, to be a force multiplier for administrative work. Estimates predict that 65 per cent of human resources rules-based processes can be automated using a combination of AI and robotics. As Navy is inundated with documents and correspondence, AI can have a direct impact on their efficiency.

AI is already generating written reports for news agencies. News outlets such as the Associated Press (AP), Fox, and Yahoo use platforms from Automated Insights to write stories about earnings reports and sports recaps. The AI-writing platform, called Wordsmith, is a natural-language-generation platform that turns data into written comprehensive text. The programme allowed the AP to publish 12 times more stories in a specific topic area with fewer errors and greater efficiency.

Since AI can learn to standardise documents, accomplish repeated tasks, and analyse data faster than humans can, it is suited to support perfectly the administration functions of the Navy.

The list of possibilities for implementation by Navy is endless: processing command check-ins and checkouts, facilitating awards write-ups and processing, executing search and creation of policies and orders, authorising travel, routing lists, disseminating white papers, and many more.

Personal Productivity

Increases in the daily personal productivity of Naval personnel can trim manhour requirements and generate efficiency at the individual level. Existing AI platforms can organise, write, and disseminate correspondence for their users. Advances in speech recognition also hold enormous promise for personal productivity. Speech-recognition capabilities can be used for authentication, instructions, planning, production, and coordination. Hours spent creating and editing documents, approving forms, passing documents and sharing information could be reduced each day.

These additional free hours allow for greater productivity at the individual level and increased opportunity for responsibility at the unit level. At the most positive extreme, never again would a sailor be refused the opportunity to attend a training school or advanced instructional course because his or her presence at work was indispensable.

The AI, which had been tracking that individual’s work for months, would be able to slide into his or her place for the duration of the absence.

Planning

Many leaders at all levels of command express frustration over lack of coordination between departments, inadequate duty turnover, and loss of long-term knowledge when vital personnel retire or redeploy. AI offers a solution through robust search function. Google has nearly perfected search techniques using a process called RankBrain.

The AI remembers what other users asked for before and the eventual end locations of their searches. It then applies that knowledge to the next search having similar inputs.

Leadership principles across the services call for the one-third–two-thirds rule of planning: one-third of the time for the leader, two-thirds of the time for the subordinate units. AI could transform this into a one-tenth–nine-tenths rule. Such capabilities could save millions of dollars from the Navy budget by determining inefficiencies, identifying discrepancies, managing accounts, and providing more-exact financial estimates for mission planning.

The most difficult step of military planning—orders development—could be streamlined with AI. Writing hundred-page documents with limitless number of annexes, appendices, and tabs has plagued staff officers since at least the Byzantine Empire in the tenth century. In only three weeks, Booz Allen Hamilton (BAH) developed its prototype Tabletop Commander programme, which can process an entire operations order and convert it into a “visually pleasing, realistic” interface for the recipient to use. The service could be helpful for amphibious ready groups (ARGs) and Marine expeditionary units (MEUs) operating in constantly changing environments.

Logistics

Transportation, logistics, and supply capabilities stand to benefit from advances in AI. Google Maps provides the basic example of the harnessed power of AI by using location and transportation data from thousands of smartphones to plan optimal transport routes. Uber uses such programmes to determine the most-exact arrival times, travel times, and pickup locations. Commercial and logistics aircraft harness AI for use in mechanical processes such as autopilot and route planning, to mitigate disruptions. In the logistics industry, courier and parcel company DHL Express and IBM collaborated on a project exploring the current and future uses of AI.

AI and logistics are natural partners since “the network-based nature of the industry provides a natural framework for implementing and scaling AI,” which helps amplify “the human components of highly organised global supply chains. The Naval Forces could use these capabilities for monitoring vehicles, fleets, buildings, bases, and operational areas.

Crisis Response

AI systems have been are helping organisations respond to humanitarian disasters and emergencies. According to the U.S. National Oceanic and Atmospheric Administration, 15 events caused over US$22 billion in damage within the United States alone in 2017. Even nominal increases in disaster response efficiency can result in immense benefits.

AI systems are used to collect data in time sequences to track changes in disaster-stricken areas for generating damage claims and publishing images for media outlets and first responders to use.

Artificial Intelligence for Digital Response (AIDR) won the 2015 Open Source Software System Challenge for its application of AI to emergencies and humanitarian crises.

The AIDR platform uses AI to sort through and categorise thousands of social media messages per minute into different categories for action, such as medical needs or sheltering. AIDR can help disaster-relief managers direct their efforts to the areas most desperately in need of aid, as well as to apply the correct types of aid (e.g., food supplies, medical assistance, heavy lift via helicopters) to the areas where they are required.

Often, in times of disaster when public services and infrastructure fail, social media platforms are the most accurate and answerable form of information for on-the-ground aid workers.

The ability of AI to sort through thousands of social media videos, pictures, and posts helps response teams map out disaster sites, provide early warnings of new disasters, and verify reports in real time.

In military terms, the combination of AI, social media, and drones could create a “common and complete picture for emergency operations centres,” aiding a commander’s command-and-control capabilities.

Training

Military training evaluations are notorious for their lack of reality. AI capabilities for the creation of data—photographs, videos, written text, and three-dimensional displays can magnify the efforts of existing small opposing forces (OPFOR) and simulated enemy forces often called Red Cell sections.

Training events involving communication, such as calls for fire, close air support and casualty evacuation requests can be challenging. Realistic and immersive decision-making exercises, information-processing evolutions, and instructional methods can be created and refined whenever needed. An OPFOR AI learns when students begin to exhibit predictable patterns and where continuous mistakes are made, and refines the training evolution to address those problems.

WalkMe, a software training platform, uses AI to develop customised learning plans for users to take advantage of their talents and learning styles. The system guides the learner through the new software and adapts the speed and depth of instruction to the learner’s abilities.

Such programmes could be used to enhance professional education and training such as distance and resident professional military education programmes and military occupational specialty (MOS) training.

Intelligence

Intelligence-collection systems are overwhelming institutional capacity for sorting and analysis. Future intelligence operations will provide even more data points from which it will be progressively difficult to “discern the truth.” The current defence Intelligence Community collects more data in one day than its entire workforce ever could analyse.

Military deception will become easier for both allies and adversaries. In the realm of intelligence collection, much research and work already have focused on new “swarm” techniques to eliminate an adversary’s ability to hide.

Ultracheap 3D-printed mini-drones could allow the United States to field billions—yes, billions—of tiny, insect-like drones.

AI assistance to intelligence can be applied to tracking and targeting. The AI exploits approximately 1.5 million different images and is adapting its system to document conservation details in categories such as geographic regions and environmental impacts. AI has led to growth in imagery analysis.

Just as AI can be programmed to learn from tax documents and contractual agreements to sort out details, the same AI can be programmed to sort through operations orders, databases, mission briefs, status-of-forces agreements, DoD policies, planning doctrine, and historical records to generate ideas and propose courses of action for commanders.

Force Protection

AI centred computer-vision technology could be used for security and base access. Providing base security constitutes a major use of manpower on naval installations. AI could augment gate-security guards by providing approval or denial for both vehicles and individuals.

The system could assess the risk applicable to new persons requesting access, determine access approvals, and increase security measures on an installation as needed, depending on internal and external threats. Photo and speech recognition can provide an additional security layer when deciding on access.

The Transportation Security Administration is implementing AI to improve the effectiveness of screening methods in U.S. airports significantly.

Mobile and roaming robotic collection platforms can also harness AI for force protection.

Personnel management

AI has innumerable applications for force structure and personnel management. For recruiting and force preservation, AI could provide early warning about at-risk personnel. Fama in U.S. is an AI-based company that screens public personas on social media platforms to detect violent or racist tendencies. An estimated 43 per cent of private employers screen potential candidates’ and employees’ social media accounts for such traits.

In the Navy, this kind of system could be used for force-preservation, recruiting, and transition programmes. The system could be used for intelligence purposes, to identify key nodes and leaders of networks.

AI not only could decrease the number of personnel necessary but also reduce costs and help recruiters to be more efficient by targeting strategically chosen candidates.

Implementation

Successful implementation of AI into the Navy cannot be outsourced totally. AI systems and the databases they use are specific to the institutions that incorporate them. AI can reach its maximum effectiveness only when the right people are paired with the right data. For successful implementation and growth, the Navy needs military specialists with knowledge of AI, human-AI collaboration, AI-database interworking, AI ethics and policy, and specific subcategories of AI, such as machine learning and deep learning.

Database compilation is a long-term process that should begin today. Fortunately, many tools and processes are available and can be replicated.

Army researchers have found that their needs for AI, such as autonomous convoys in rough environments, manned-unmanned teams for ISR targeting, and intelligence analysis, are not yet “of significant interest” to private companies. Ultimately, these business applications provide not a specific solution but rather inspiration and a roadmap that the Department of Navy (DoN) can use to implement AI strategically throughout its departments.

Reference Text/Photo:

Heller, Christian H. (2019) “The Future Navy—Near-Term Applications of Artificial Intelligence,” Naval War College Review: Vol. 72 : No. 4 , Article 7.

www.lockheedmartin.com

www.northropgrumman.com

www.navy.mil

Comments () Views (39796 ) Print

Send To Friend