When forward-deployed Army soldiers need air support, an operations center is tasked with identifying and assigning aircraft aid. With traditional software, an operator moves through a multistep process to search for available aircraft, identify their call signs and assess the munitions they carry. Pulling this relevant information can take several minutes—a long time to wait when making “real-time” decisions for immediate support.
Considering the massive amount of information the U.S. Department of Defense must sift through every day and increasingly sophisticated UAVs and UASs collecting even more data, it’s no surprise the Pentagon has turned to artificial intelligence for help.
The newly launched Chief Digital and Artificial Intelligence Office and efforts and strategies such as the Artificial Intelligence and Data Acceleration initiative, Joint All Domain Command and Control and JAIC shows that DoD recognizes the potential of AI in decision compression.
The integration of AI is still largely siloed because access to networks that support individual missions remain stovepiped, as does the data access within those enclaves. Although efforts like JADC2 look to address this issue, difficulties associated with standing up hybridized infrastructures and training operational mission machine learning models keeps DoD from making significant progress.
The Pentagon also lacks standardized authorization to operate for automation and ML, further constraining the training and deployment of these models and degrading the capability of AI tools supporting the warfighter.
Driving these issues is the perception of AI as a separate add-on tool. Instead, it should be considered a core element of technology and part of an agency’s foundational IT infrastructure—whether on the ground, in the cloud or through hybrid systems.
AI can no longer be treated as an R&D project. As agencies adhere to fundamental development processes as part of the Information Technology Infrastructure Library and deploy infrastructure in a more agile manner, they must also include the fundamental principles of MLOps and DataOps at the beginning of the planning process.
By integrating AI into the planning process, agencies can partially mitigate the lack of standardized ATO, as officials will have time to identify and authorize the necessary individuals as a part of the planning process.
But even beyond the planning process, practical AI implementation must start at the most fundamental level - quality data and a robust data governance process. Without clean, easy-to-understand data, AI cannot provide effective analysis for quick and correct decisions.
“Good” data is especially critical for DoD, as every delay or decision based on insufficient or inaccurate information can potentially cost a life.
By ensuring data is streamlined and gathered from various sources, including Human intelligence (HUMINT) on the ground and in the field, signals intelligence (SIGINT) ground and airborne sensors, cameras and social media, DoD can ensure that any AI implemented into their system is effective and accurate.
With clean data, the DoD can look to the main benefit of AI in defense - accelerating the observe–orient–decide–act loop.
AI’s role in Defense
The OODA loop is a process utilized by leaders and managers to make quick decisions based on the information available at any given time, keeping the human involved in the process. Although full automation may be the end goal for some missions, checkpoints must be in place to maintain an appropriate chain of command, and humans must have final authority due to the non-linear nature of war.
Over time, as data expands and ML advances, AI continues to evolve with the needs of individual missions.
For example, some missions may need AI to scan hundreds of satellite images looking for a specific tank, while others may require AI to urgently translate several messages for human review. These specific mission parameters lend themselves to non-linear warfare. AI can advance the OODA loop by quickly consolidating more information than humans can review, leading to more accurate and timely decisions.
Some may argue that because of this ever-shifting state of warfare, using AI models to accelerate the OODA loop may complicate the battlespace and create unnecessary bias, having strategic ramifications and long-term ripple effects.
However, the reality is humans make decisions based on memory and experience and warfare is more than probability, statistics and heuristics. Combining human experiences with AI’s capacity to quickly generate scenarios can allow for safer, more appropriate decisions.
AI’s Lasting Impact
Leveraged appropriately, AI can scale the impact of additional experiences created based on all data it has access to, amplifying warfighters’ situational awareness and scaling their capacity to maintain superiority over the enemy.
The key to success is keeping the human at the center of AI design and avoiding technology innovations that will advance faster than a fighter’s ability to make decisions.
To truly benefit from AI’s value, we need to open the aperture. We must look beyond tactically focused implementations like AI-driven weapons systems and build warfighter-centered AI solutions that deliver relevant information faster.
Dr. Allen Badeau is the chief technology officer for Empower AI, as well as the director of the Empower AI Center for Rapid Engagement and Agile Technology Exchange (CREATE) Lab.
Have an Opinion?
This article is an Op-Ed and the opinions expressed are those of the author. If you would like to respond, or have an editorial of your own you would like to submit, please email C4ISRNET Senior Managing Editor Cary O’Reilly.