Artificial Intelligence

This is an adaptation of an essay By Ensign Clarisse Joy Absalon, U.S. Navy, July 2025 Proceedings Vol. 151/7/1,469, U.S. Naval Academy Class of 2025 Capstone Essay Contest Winner

Introduction

Under the US National Artificial Intelligence Initiative Act of 2020, AI is defined as “a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments.”1 AI is thus integrated more as a supporting resource than as a replacement for human labour or decision-making.

AI’s potential advantages in submarine operations need to be weighed against the possibility that excessive reliance on it will weaken cognitive acumen. Some preliminary research suggests that, with careful management, the benefits can outweigh the risks. Specifically, three areas could benefit when exploring AI integration in submarine operations—maintenance prediction, sonar detection, and decision-making.

Maintenance Prediction

In an environment that requires constant monitoring and system evaluations, AI could prove especially useful in predicting maintenance by overcoming human performance factors, (eg time of day or night, distractions, and fatigue). Through reinforcement training, during which AI programs learn through trial and error, developing algorithms that can more quickly predict issues and recommend solutions without human involvement.

The USN Seawolf ship-control system (SCS) is a relevant example of an automated performance and maintenance monitoring system. The SCS’s performance monitoring/fault localization feature efficiently detects and reports any SCS component failures. Furthermore, using its input/output debug monitoring software, engineers can more easily analyse and troubleshoot issues because of the clearer digital visual and centralized data.

AI could enhance the SCS’s ability to predict maintenance issues by tracking maintenance patterns and the age of ship components. One of the most basic forms of machine learning is supervised learning, in which algorithms are trained to classify data and predict outcomes through inputted examples.

By applying supervised learning, predictive maintenance schedules could be developed and refined with more accuracy. Information collected by the AI model also could be used to support training simulations. By automating and predicting maintenance issues, AI would allow maintenance checks to be more efficiently prioritised and manpower reduced.

Sonar Detection

Significant assistance to submarine sonar operators could be gained from AI. Sonar operators are responsible for the constant surveillance, of the acoustic environment: the concentration required to filter and fuse the acoustic data potentially associated with targets of interest, separate from the background noise of the ocean, is extremely demanding and skilled.

Detecting and refining the acoustic contact picture for the submarine, must then be filtered, fused and classified, the tracks managed, localised (and perhaps engaged). Their work requires continuous monitoring and concentration. It can be extremely fatiguing.

An AI sonar-detection algorithm could reduce the workload by recognizing patterns of marine life, weather generated wave noise, underwater terrain features, and known friendly underwater objects (unmanned underwater vehicles, underwater mines, etc.). The tool also could enhance the acoustic skills associated with detection, filtering, fusion, classification, tracking and localisation of contacts of interest.

Great care must be taken to understand that AI is an enhancement to, and not a replacement of, Sonar operator / command classification acoustic skills.

Decision-Making

The demanding roles, suboptimal sleep schedules, and dangerous operating environment, require quick, clear and accurate decisions for all submarine operations.

AI decision-making aids could enhance optimal decision-making path. From a small amount of data, AI can produce an extensive amount of synthetic information that it can then use to simulate possible real-life situations. This can assist the submarine command team to more rapidly and comprehensively analyse situations.

Further, AI decision-enhancing algorithms could tailor recommendations to specific missions. AI provides a consistent performance 24/7 ; human performance will vary based on several factors, including time of day or night, distractions, and fatigue. Given a database of incidents, such as the USS San Francisco (SSN-711) or USS Connecticut (SSN-22) seamount collisions, an AI algorithm could be used to predict operational/navigational risk, generate incident-prevention measures and analyse the risk associated with potential decisions.

AI algorithms could also simulate previous training evolutions and mishaps and assist the training of junior officers.

Mitigating Risk

AI algorithms and Machine Learning tools present clear opportunities to enhance submarine operations as indicated above (and, potentially, other areas of submarine operations). The risk remains that, although AI tools can streamline/enhance manual and mental procedures, such reliance would result in a deviation from the questioning ‘what if?’ tactical alertness so critical to past, present and future submarine operations.

As demonstrated by the San Francisco navigational collision incident, deviation from standards and failure to adhere to procedures will always be a risk that no AI algorithm can remove. AI algorithms can mitigate, but not eliminate, human error. For the foreseeable future, AI will require human interaction to offer reliable enhancement.

Summary

The automated processes Machine Tool learning and generative models of AI algorithms, indicated above, could enhance submarine operational efficiency, improve maintenance efficiency and upgrade training aids.

AI algorithms can mitigate, but not eliminate, human error. For the foreseeable future, AI will require human interaction to offer reliable enhancement.