ChatGPT: The Future “Loyal Wingman” of Navy Paperwork?
The MOC
Then-Captain Hopper at work. Photo from Deconnection.org.
By
Dr. Steven Wills | Mavreen Smiel
August 1, 2023
Admiral Grace Hopper — arguably the founding parent of modern computer science — once said, in the 1980’s, that the world was at best in the “Model T” era of computing and that, “You ain’t seen nothing yet,” in terms of computer power. The word of advanced digital processing has traveled much further since Hopper’s time. It has now reached the stage where artificial intelligence (“AI”) is speaking for itself in terms of what it has and can learn from human feedback. ChatGPT, one of the most recent such products is a conversational / interactive AI tool that in the words of its creators, “Can follow complex instructions in natural language and solve difficult problems with accuracy.”
Many are concerned because such a tool has been used by students to cheat while writing term papers or replace skilled human writers. However, there is also exciting potential for ChatGPT to reduce the burden of mundane data entry work — the lot of so many in the workforce. ChatGPT has the potential to revolutionize what being a commissioned or non-commissioned officer means in terms of work/life balance and professional accomplishment. A navy.mil ChatGPT server accessible from any ship or shore station has the potential to take over much of the mundane and demanding naval paperwork that increasingly dominates the working day of Navy people. Freed from the burden of such administrative tasks, naval leaders can focus on training to be better operators and warfighters, taking care of the people under their command, and achieving a better work/life balance with more time off the ship and out of the office.
The Rising Paperwork Load Afloat
Operational and administrative paperwork have always been part of naval officer duties. In the past several decades, however, with the advent of computer word processing and automated data transfer systems in the fleet, those paperwork responsibilities have exponentially grown to take up more of the average officer’s time. These reports reference equipment maintenance and repair, training, supply and ship logistics, ammunition expenditure, navigation, and crew medical status among the many categories of report. These are usually drafted for electronic transmission via ships’ digital communication equipment.
Senior enlisted personnel and junior officers usually originate much of it in the course of their duties and often must push it though successive levels of the chain of command to get it released by the commanding officer. There is often a time or operational requirement to submit, and many report drafters have labored long after the ship’s regular working hours have ended, to complete and send these reports.
Ironically, the development of advanced data processing equipment and shipboard computer networks designed to empower the crew have instead increased their administrative burden. Despite numerous attempts by the Navy to reduce the administrative burden, the overall requirement for reporting seems to grow, according to many. Increased time devoted to administrative affairs has been cited by many junior officers as detracting from their ability to take care of their people and train to be better warfighters. Some officers leaving the Navy for the civilian workforce have cited the service’s vast paperwork burden as one of their key reasons for leaving, especially when it takes up time in port better spent with family and friends off the ship.
How Does ChatGPT Work?
It was not until the mid-20th century that hardware had become powerful and sophisticated enough to make AI a reality. During this time, machine learning (“ML”) emerged as a practical application of AI. Today, machine learning has two main goals: to create and improve models by classifying data and to make predictions or decisions based on those models.
But what is a machine learning model? Every model, regardless of size or complexity, is nothing more than a mathematical equation or algorithm. Machine learning algorithms differ from conventional ones because they are created using sample data, called “training sets,” without being explicitly programmed. Although many models perform specialized functions, Large Language Models (“LLMs”), such as ChatGPT, mimic human language to accomplish a variety of tasks, like drafting a report. Training and classification sets require maintenance and careful curation by engineers. Bad data can have disastrous consequences, many of which are not always readily apparent.
How can ChatGPT Potentially help the Navy’s Paperwork Morass?
As mentioned above, ChatGPT and other LLMs excel at generating text. A Navy-centric ChatGPT model could assist with a variety of reports — from Equipment Casualty reports (“CASREPs”) to enlisted performance evaluations and officer fitness reports — while side-stepping many of the issues with civilian models have. Additionally, it would have an offline mode for security and for preserving bandwidth while underway.
Before heading to sea, the model could determine the ten most common issues for each system on a ship and save appropriate responses. These could either be used as-is or be a starting point for refinement. For personnel attached to the ship, copies of their data are moved to the ship’s systems (called “local copies”). When a form or report is needed, the model can pull data necessary from the ship’s local copy. This is done through Application Programming Interfaces (“APIs”). APIs allow programs to send and receive data from a variety of sources in a standardized, automated way. By pairing models, APIs, and automation, forms could be pre-filled with appropriate data while allowing the model to generate narrative sections of the report. This would not only save an individual’s time but reduce or eliminate errors which reduces processing times. When the ship returns home or arrives in a secure location with better bandwidth, changes can be synced, and updates can be retrieved through an API call.
ChatGPT and Machine Learning at a Systemic Level
While these models would certainly save countless hours of work, the greatest benefits would come from infrastructure improvements needed to use it in the first place. Robust modern web, software, and data architectures will allow AI to become the keystones in a comprehensive suite of capabilities to address the Navy’s current and future administrative needs.
For example, Ammunition Transaction Reports (“ATRs”) do not need ChatGPT functionality at all. Instead, a simple web-like user interface will display information about the ship’s ammunition that is stored in a Structured Querry Language (“SQL”) database (with backups). When ammunition is expended and numbers are updated, a small program will update the Conventional Ammunition Integrated Management System (“CAIMS”) database via an API call. This API message will have all the necessary data pre-formatted, eliminating additional points of human error and the need for a routine message. Robust event logging with timestamps creates auditable trails automatically.
In all these cases, the real power of AI/ML can be brought to bear stateside in secure infrastructure. Anomaly detection models can reveal potential issues — or advantages — while predictive models can provide better forecasting on real-time data flowing in. Those models could optimize parts procurement, scheduling, and maintenance periods, shortening time spent in the shipyard or general maintenance activity.
For personnel issues — such as reimbursements for travel — requests that fall within established guidelines could be automatically approved, saving more complex or sensitive tasks for human review. Over time, the system could identify unused or redundant processes paving the way for greater efficiencies. All these insights can be surfaced to leaders through dashboards, replacing a patchwork of proprietary systems, spreadsheets, and emails — many of which need to be manually created or updated.
If all of this sounds too good to be true, it is not. This functionality is well within the bounds of current technology. While the Navy has many pressing concerns, each competing for its precious resources, the Navy can no longer afford to slip further behind in technology. The time has come for the Navy to work smarter and get Sailors back to doing what Sailors should be doing: training to fight and win.
Dr. Steven Wills is the Navalist for the Center for Maritime Strategy.
Mavreen Smiel is a Senior Software Engineer in Test with more than a decade of experience in automation frameworks, real-time data delivery systems, and AI/ML infrastructure.
The views expressed in this piece are the sole opinions of the author and do not necessarily reflect those of the Center for Maritime Strategy or other institutions listed.
Center for Maritime Strategy
By Dr. Steven Wills | Mavreen Smiel
Admiral Grace Hopper — arguably the founding parent of modern computer science — once said, in the 1980’s, that the world was at best in the “Model T” era of computing and that, “You ain’t seen nothing yet,” in terms of computer power. The word of advanced digital processing has traveled much further since Hopper’s time. It has now reached the stage where artificial intelligence (“AI”) is speaking for itself in terms of what it has and can learn from human feedback. ChatGPT, one of the most recent such products is a conversational / interactive AI tool that in the words of its creators, “Can follow complex instructions in natural language and solve difficult problems with accuracy.”
Many are concerned because such a tool has been used by students to cheat while writing term papers or replace skilled human writers. However, there is also exciting potential for ChatGPT to reduce the burden of mundane data entry work — the lot of so many in the workforce. ChatGPT has the potential to revolutionize what being a commissioned or non-commissioned officer means in terms of work/life balance and professional accomplishment. A navy.mil ChatGPT server accessible from any ship or shore station has the potential to take over much of the mundane and demanding naval paperwork that increasingly dominates the working day of Navy people. Freed from the burden of such administrative tasks, naval leaders can focus on training to be better operators and warfighters, taking care of the people under their command, and achieving a better work/life balance with more time off the ship and out of the office.
The Rising Paperwork Load Afloat
Operational and administrative paperwork have always been part of naval officer duties. In the past several decades, however, with the advent of computer word processing and automated data transfer systems in the fleet, those paperwork responsibilities have exponentially grown to take up more of the average officer’s time. These reports reference equipment maintenance and repair, training, supply and ship logistics, ammunition expenditure, navigation, and crew medical status among the many categories of report. These are usually drafted for electronic transmission via ships’ digital communication equipment.
Senior enlisted personnel and junior officers usually originate much of it in the course of their duties and often must push it though successive levels of the chain of command to get it released by the commanding officer. There is often a time or operational requirement to submit, and many report drafters have labored long after the ship’s regular working hours have ended, to complete and send these reports.
Ironically, the development of advanced data processing equipment and shipboard computer networks designed to empower the crew have instead increased their administrative burden. Despite numerous attempts by the Navy to reduce the administrative burden, the overall requirement for reporting seems to grow, according to many. Increased time devoted to administrative affairs has been cited by many junior officers as detracting from their ability to take care of their people and train to be better warfighters. Some officers leaving the Navy for the civilian workforce have cited the service’s vast paperwork burden as one of their key reasons for leaving, especially when it takes up time in port better spent with family and friends off the ship.
How Does ChatGPT Work?
It was not until the mid-20th century that hardware had become powerful and sophisticated enough to make AI a reality. During this time, machine learning (“ML”) emerged as a practical application of AI. Today, machine learning has two main goals: to create and improve models by classifying data and to make predictions or decisions based on those models.
But what is a machine learning model? Every model, regardless of size or complexity, is nothing more than a mathematical equation or algorithm. Machine learning algorithms differ from conventional ones because they are created using sample data, called “training sets,” without being explicitly programmed. Although many models perform specialized functions, Large Language Models (“LLMs”), such as ChatGPT, mimic human language to accomplish a variety of tasks, like drafting a report. Training and classification sets require maintenance and careful curation by engineers. Bad data can have disastrous consequences, many of which are not always readily apparent.
How can ChatGPT Potentially help the Navy’s Paperwork Morass?
As mentioned above, ChatGPT and other LLMs excel at generating text. A Navy-centric ChatGPT model could assist with a variety of reports — from Equipment Casualty reports (“CASREPs”) to enlisted performance evaluations and officer fitness reports — while side-stepping many of the issues with civilian models have. Additionally, it would have an offline mode for security and for preserving bandwidth while underway.
Before heading to sea, the model could determine the ten most common issues for each system on a ship and save appropriate responses. These could either be used as-is or be a starting point for refinement. For personnel attached to the ship, copies of their data are moved to the ship’s systems (called “local copies”). When a form or report is needed, the model can pull data necessary from the ship’s local copy. This is done through Application Programming Interfaces (“APIs”). APIs allow programs to send and receive data from a variety of sources in a standardized, automated way. By pairing models, APIs, and automation, forms could be pre-filled with appropriate data while allowing the model to generate narrative sections of the report. This would not only save an individual’s time but reduce or eliminate errors which reduces processing times. When the ship returns home or arrives in a secure location with better bandwidth, changes can be synced, and updates can be retrieved through an API call.
ChatGPT and Machine Learning at a Systemic Level
While these models would certainly save countless hours of work, the greatest benefits would come from infrastructure improvements needed to use it in the first place. Robust modern web, software, and data architectures will allow AI to become the keystones in a comprehensive suite of capabilities to address the Navy’s current and future administrative needs.
For example, Ammunition Transaction Reports (“ATRs”) do not need ChatGPT functionality at all. Instead, a simple web-like user interface will display information about the ship’s ammunition that is stored in a Structured Querry Language (“SQL”) database (with backups). When ammunition is expended and numbers are updated, a small program will update the Conventional Ammunition Integrated Management System (“CAIMS”) database via an API call. This API message will have all the necessary data pre-formatted, eliminating additional points of human error and the need for a routine message. Robust event logging with timestamps creates auditable trails automatically.
In all these cases, the real power of AI/ML can be brought to bear stateside in secure infrastructure. Anomaly detection models can reveal potential issues — or advantages — while predictive models can provide better forecasting on real-time data flowing in. Those models could optimize parts procurement, scheduling, and maintenance periods, shortening time spent in the shipyard or general maintenance activity.
For personnel issues — such as reimbursements for travel — requests that fall within established guidelines could be automatically approved, saving more complex or sensitive tasks for human review. Over time, the system could identify unused or redundant processes paving the way for greater efficiencies. All these insights can be surfaced to leaders through dashboards, replacing a patchwork of proprietary systems, spreadsheets, and emails — many of which need to be manually created or updated.
If all of this sounds too good to be true, it is not. This functionality is well within the bounds of current technology. While the Navy has many pressing concerns, each competing for its precious resources, the Navy can no longer afford to slip further behind in technology. The time has come for the Navy to work smarter and get Sailors back to doing what Sailors should be doing: training to fight and win.
Dr. Steven Wills is the Navalist for the Center for Maritime Strategy.
Mavreen Smiel is a Senior Software Engineer in Test with more than a decade of experience in automation frameworks, real-time data delivery systems, and AI/ML infrastructure.
The views expressed in this piece are the sole opinions of the author and do not necessarily reflect those of the Center for Maritime Strategy or other institutions listed.