Artificial intelligence (AI) and its related technologies promise to increase the speed, efficiency, and effectiveness of each phase in the intelligence cycle to better support naval commanders.
AI holds promise for every military function, but especially for intelligence because of the availability of large datasets, an intelligent workforce, the proliferation and integration of technology, and repeatable processes widely understood by a diverse workforce.
AI and the Intelligence Process
Joint Publication 2-0 articulates the six categories of intelligence operations as planning and direction, collection, processing and exploitation, analysis and production, dissemination and integration, and evaluation and feedback. Numerous examples from research, private industry, and existing technologies showcase AI’s ability to improve naval intelligence.
Planning and Direction
AI tools can help planners work faster and more efficiently. Integrating operational intelligence components with national assets is difficult in current structures. Many tactical and theater platforms do not link to national intelligence databases, and satellite architecture for transmitting information is lacking. AI can help these networks communicate and allow planners, collection managers, and analysts to know what intelligence already exists and what collection assets are available to task.
Planning also involves writing operations orders and directives. AI-based advances in speech-to-text programs and basic reading comprehension programs can generate reports, summaries, and transcripts for rapid dissemination and consumption. AI can also ensure the most relevant instruction reaches the required target in the most efficient manner.
AI also has the potential to train leaders on the wide variety of intelligence topics that staff officers are required to understand. Every type of intelligence collection is teeming with legal and technical know-how. Analysis is a distinct subdiscipline and regional expertise is a prerequisite to make any valuable assessment on a topic. AI-supported wargames, such as those used at Marine Corp University, can provide the training needed to create superior analysts and officers in a shorter period of time than traditional experience.
Collection
Intelligence collection stands to benefit substantially from AI. An increase in internet users and networked systems means that collection opportunities have significantly increased. The amount of information created from improved collection methods amounts to levels of reporting that analysts no longer can evaluate without automated assistance. AI can organize and manage this collection in ways humans cannot. Similar to advertising recommendations on web pages or flash-sale notifications from email or Twitter, collection managers could be instantly notified of excess assets for tasking as they become available.
AI can increase the speed and endurance of collection systems far beyond the proficiency of human operators. For example, researchers supporting the National Geospatial-Intelligence Agency (NGA) with AI technologies identified surface-to-air missile sites 80 times more efficiently than traditional human searches. The combination of AI platforms with robotic swarm technology could enable new collection techniques such as “wide-area, long-persistence, surveillance.”
For environmental intelligence needs, mass information collected from social media, cell phones, and satellites already helps researchers acquire never-before-used socioeconomic data. This information is critical for stability operations, such as those in Iraq and Afghanistan. New satellite networks report on changing weather and terrain conditions that have obvious military benefits. Underwater and surface-based sensors can operate persistently from internal or solar power to collect real-time and multi-dimensional data, detect movement, and perceive tectonic shifts. In addition, AI already has proven its ability to plan routes, road trips, and detours based on available data, capabilities which can support intelligence preparation of the battlefield (IPB) requirements. This battlespace information will be vital for commanders in a future military environment characterized by dense urban areas, electrical grids, overlapping logistics routes, and complex communications infrastructure.
Processing and Exploitation
AI will support processing and exploitation by its ability to sort, organize, and identify trends in mass amounts of data, which could require thousands of man-hours to review. By doing so, AI will enable analysts to discover and manipulate data in innovative ways. Intelligence applications today use sensors feeding into distinct databases on unique networks and accessible from specific applications. The result is such that analysts’ current ability to use all available sources of information “is severely restricted by an inability to understand what we have already got.” AI helps overcomes those constraints.
Analysis and Production
Analysts at RAND point out the failure of the current stovepiped intelligence system and its impact on analytical procedures:
“Add to the volume of data collected by national technical means the even greater volume of data available through commercial ISR platforms and openly available on the internet, or crowdsourced and uploaded as needed by billions of smartphones and other sensors around the planet—and it becomes even more obvious that analysts, let alone commanders and decision makers, do not have access to all available sources of information.”
AI-based networks can overcome this gap by linking various tactical and strategic intelligence structures to correlate data and derive insights from the widest sets of information. This cross-cueing and comparison enables better analysis by not only increasing the amount of information available, but also by enabling intelligence sections to determine the reliability and credibility of reporting. Big data and AI can free analysts from the routine duties of searching and compiling reports, allowing them the time to concentrate on analytical tradecraft.
Production in the intelligence community requires analysts to present information in a usable way. Creating products is manpower-intensive. AI systems can generate the same information in multiple different forms for various consumers. Going further, if AI allows analysts to access larger amounts of distant information, a single intelligence center could support operations in multiple different environments simultaneously. Soon, geographic or unit distinctions may not matter. To create this reality, the Navy will need to invest in communications infrastructure to increase the bandwidth and network capabilities on board vessels operating far from shore.
Dissemination and Integration
AI will enable finished intelligence to be disseminated faster and wider than current organizational models and technical capabilities allow. Email filters in programs like Gmail already determine which information needs to be viewed first, and could be expanded for wider applications. In another example, neurolinguistic programming (NLP) has been extensively developed by companies such as Amazon to interact with humans using the spoken word. NLP can be used to pass commands, change products from one system to another, or translate products for wider coalition dissemination.
Autosearch and monitoring software can identify information that needs to be disseminated immediately. Facebook uses similar AI to detect suicidal ideations in its users and notify human moderators of the posts, who can then contact the relevant authorities when necessary. This technology can be used to maintain constant watch officer-style awareness on regional or global reporting to identify and distribute important intelligence in real time.
Evaluation and Feedback
AI can support the evaluation of intelligence and identify and fix mistakes in processes. AI systems adapt and learn from previous decisions in ways that rules-based programs and humans do not. AI systems can analyze system performance and provide feedback to improve decision making, minimize risk, and offer new insights that human observers might fail to notice. In addition, this learning capability helps software avoid becoming obsolete and respond to a wider variety of tasks at a lower fiscal cost.
Developing an AI Community
AI development will require close cooperation between the Navy and private industry. Silicon Valley and other tech leaders are the primary developers of AI-technologies and appear poised to remain so for the coming years. Naval intelligence leaders’ primary impact with AI will not be from driving research projects in new directions, but from learning about and implementing commercial systems for military applications.
The Navy needs to educate leaders about AI to build trust in these systems. If commanders do not trust their systems, it will slow down decision making to human speed and negate any advantage over the adversary.
Developing innovative systems to manage and apply big data will challenge the Navy’s leaders. New policy should cover not only the proper uses of AI for intelligence, but the information technology (IT) infrastructure, organization, and limitations to facilitate the use of these systems. AI capabilities will require new analytical and collection tradecraft based on the best practices of human-machine teaming. One proposed model argues for the adoption of data acquisition teams to act as digital collection elements, data curation teams to assess information’s veracity, data exploitation teams to monitor algorithms and computer models, and data visualization and distribution teams to properly produce and disseminate information.
New military specialties will be required for an AI-enhanced intelligence community. Data scientists, machine learning engineers, big data specialists, and software engineers may all be required in the Navy’s future. Private industrywill require an estimated 140,000 to 190,000 specialists for advanced AI-related analytical positions, in addition to another 1.5 million “data-savvy managers.” Attracting these specialists away from high-paying commercial jobs will strain Navy recruiting. One possibility is to focus on the post-military job skills the Navy can offer. If Navy training pipelines can turn young recruits into data science and computer engineering specialists in a four-year enlistment, the service will have a major incentive to entice young Americans to delay college or forego the private sector.
Approximately 50 percent of current workplace activities could be automated with existing technologies. Machine learning systems are improving six fold every year, and the world creates 2.2 billion gigabytes of data every day. The Navy will not be immune to these disruptions and needs to adapt for the future. The intelligence community is an ideal area to begin the transition to human-machine teaming since AI can improve every step in the intelligence cycle. While the adoption of these technologies will not be easy, an educated workforce, the proliferation of technology and databases, and standardized and repeatable processes make AI’s implementation worthwhile. Doing so will give naval leaders and the naval intelligence community the advantages they need to win during 21st century warfare.