Businesses venturing into AI could learn from this DOD project

Businesses looking for ways to kick-start artificial intelligence projects can look for guidance to an innovative project being planned by the U.S. Department of Defense.

The DOD’s Joint Artificial Intelligence Center has issued a Request for Information that outlines a project that would use AI to find patterns and anomalies in historic helicopter maintenance records in order to predict problems, improve safety, and decrease maintenance costs. Businesses already using AI technologies might want to jump on this business opportunity with the DOD (RFI responses are due May 20). But businesses new to AI can look at the DOD’s approach and goals as a model in defining their own initial AI projects to improve the odds of success.

Here are key approaches the military’s Joint AI Center is following that I see as worthy of being emulated by other AI projects. The DOD project:

Fits a defined business objective (in this case, defined in the U.S. military’s AI Strategy
Defines bite-sized wins rather than trying to solve every part of the problem domain
Uses a defined, obtainable data set for use by AI algorithms
Solves a problem not easily performed by people
Avoids user and leadership pushback by:
- Minimizing impact on current business processes
- Assisting workers with their jobs rather than raising suspicions that humans will be replaced
- Being achievable without a huge budget
- Meeting established ethical guidelines
Uses available technologies
Will define a test/verification strategy to measure results

I consider these project attributes worthy of notice because the military, by creating the Joint AI Center

For businesses considering how they can leverage artificial intelligence technologies to solve business problems or create a competitive advantage, the qualities in this DOD project proposal are the ones to emulate in early project definitions (although you can achieve major wins without saving lives). Before spending millions of dollars on huge AI projects, businesses looking for an AI entree project likely will succeed best by starting small, defining a project that increases competitiveness without having to go all-in with a particular vendor or technology, and either succeeds or fails in a short time frame with minimal costs. Early AI projects will be about learning what works and what doesn’t work for the business. It is best to learn with projects that provide feedback quickly and cheaply.

A proposed U.S. Defense Department project to use artificial technologies to improve military helicopter maintenance could serve as a model for businesses trying to figure out how to begin adopting AI for strategic advantage. The proposed project, being spearheaded by the DOD’s relatively new Joint Artificial Intelligence Center, defines exemplifiable practices to keep the project focused, agile, and maximizing the odds of showing early value.

Maybe mention this is Part 2 of an earlier project that used NLP to add metadata to maintenance records based on free-form notes added by technicians.

The project:

Fits a defined business objective (in this case, defined in the U.S. military’s AI Strategy
Defines small, incremental outcomes rather than trying to solve every part of the problem domain
Uses a well-defined, obtainable data set for use by AI algorithms
Solves a problem not easily performed by people
Avoids user and leadership pushback by:
- Minimizing impact on current business processes
- Assisting humans with their jobs rather than raising suspicions that they will be replaced
- Being achievable without a huge budget
- Meeting established ethical guidelines
Uses available technologies
Will define a test/verification strategy to measure results

This DOD project is being proposed in a Request for Information posted on the government’s award-management system. Responses to the RFI are due May 20. After evaluating responses to this RFI, the DOD will decide whether to move forward with a formal project and request proposals.

In this proposal, the DOD is looking for ways to use AI to:

Use natural language processing to improve helicopter maintenance logbook accuracy. For example, augmenting standardized check-box type of maintenance information with free-form text descriptions entered by maintenance crews
Use machine learning to predict future helicopter maintenance needs based on historical operational trends
Identify operational and maintenance anomalies that could be linked to data such as how the helicopter was flown, the weather, environment (e.g. sandy desert, salty air) that are captured from multiple sensors and tear-down records

Businesses and other government agencies can learn from this DOD proposed project for what it defines and what it does not define. The proposal is meant to use data the military services already collect and that can be gathered for data-analytic experiments, be implementable without new advances in technology, and look for anomalies and patterns in the data to improve helicopter maintenance and servicing that can save money and save lives by solving hardware failures before they happen. The proposal does not try to use AI for futuristic, whiz-bang approaches to auto-pilot helicopters in superhuman ways to defeat enemies on the battlefield or use AI to auto-fire the helicopter’s weapons in ways guaranteed to kill only the bad guys while preventing all civilian casualties.

In other words, this project proposal is realistic. It is small enough to be readily achievable without having to make major new investments and can be completed in a reasonable time frame. If the project succeeds, the military services will save money and potentially save the lives of helicopter pilots, crews, and passengers.