Army hopes AI will give soldiers an informational edge
Illustration on iStock
In recent years, the Army has introduced the concept of “information superiority,” the ability of soldiers to make decisions and act faster than their adversaries. The service now believes that artificial intelligence is the key to making strategy a reality.
The popularity and capabilities of artificial intelligence have exploded, making large language models such as ChatGPT and other AI systems more easily accessible to the general public. Many in both industry and the Pentagon are exploring potential military applications for this technology, and the Army is no exception.
Army Cyber Command Commander Lt. Gen. Maria Barrett said AI “has the highest potential to really drive change…but it poses a very real challenge for us, not just the information side.”
Maj. Gen. Matthew Easley, deputy chief of staff for information operations in the Office of the Undersecretary of Defense for Policy, said the service is “moving away from traditional information operations, how different information effects are combined to create the desired synergies in operations, to the new concept of information superiority.”
The goal of the concept is to allow the Army to “take the lead” in the information environment and “see itself, know itself, and act more quickly,” Easley said at an American Army Association event in June. The information advantage, he said, includes five main functions. Protect soldier and army information. Educate and inform domestic audiences. Inform and influence foreign audiences. and waging information warfare.
“All five of these areas can use artificial intelligence and machine learning to some effect,” he added.
Easley helped found the Army’s AI Task Force in 2019. However, during his tenure the team faced two challenges in implementing AI across the Army. The move to hybrid cloud environments and mobile devices.
The Army “will continue to have many legacy data centers, but it needs to grow fast, so it needs to move around the world. The cloud environment makes global operations much easier,” he said. The service is asking for $469 million in fiscal 2024 to migrate to the cloud and invest in its data environment, according to Army budget documents.
“You can’t have AI or machine learning without a repository of data,” Barrett said at the AUSA event. The Army Cyber Command has invested heavily in its big data platform, “doubling the amount of data flowing into the platform, the number of parsers, and the amount of storage of what we currently store,” she said. “We are continuing on that trajectory, and that means we are definitely at the stage where we are starting to take advantage of AI capabilities.”
For headquarters, AI is primarily used for network defense, but it can also be applied to the “information side,” she said. “What does it mean to bring in all kinds of disparate feeds … It allows us to really understand the information baseline of a given environment, when and how it is changing and where it is changing.
While the proliferation of mobile devices has greatly expanded the volume of potential feeds, Easley said it could also expand potential targets for attackers. These devices “have many functions, [and] It also comes with many vulnerabilities. We need to think and use artificial intelligence…not only to protect ourselves, but also to manage the vast amount of data out there. ”
And in potential conflicts, AI could help soldiers sort through all the data and get the right information on the military’s proverbial “right arrow,” Army Chief of Staff Gen. James McConville said at a media briefing in June.
For fiscal year 2024, the service is calling for $283 million in AI and machine learning, which will include research and development funding for enhanced autonomy experiments, as well as funding for AI/ML-enabled program activities for systems such as an integrated vision augmentation system, optionally a manned combat vehicle (recently redesignated as the XM30 Mechanized Infantry Combat Vehicle) (remote combat vehicles, TITAN ground stations, and “smart sensors with edge processing,” according to the Army budget). ”) documents.
“We are certainly looking at ways to leverage AI to make both new and developing capabilities more effective,” Secretary of War Christine Wormuth said at a briefing. In particular, she said the service is leveraging the AI-targeted program in her Convergence exercise project.
Project Convergence is the service’s contribution to the Department of Defense’s joint all-area command and control concept, which aims to link sensors and gunners through a network. In the final exercise at the end of 2022, participants used the Army’s Firestorm system, an artificial intelligence-powered network that combines sensors and shooters, to transmit information to Australian forces participating in the experiment, according to an Army release.
The service also uses AI for predictive logistics, McConville said. “We now have him using AI to predict what we need for parts, which is very important when you go into a large military,” he said.
Beyond simple maintenance, predictive logistics also includes the Army’s supply of different types, such as fuel and ammunition, “how we look at consumption and how we can predict how to get the right supplies to where they are needed,” said Army Assistant Secretary for Maintenance Timothy Godett.
The goal is to pre-plan where those supplies need to be sent and when maintenance needs to be done, “rather than react,” Godett said at the Defense Industry Association’s Tactical Wheeled Vehicles conference.
“Scheduled maintenance is correct, but the situation is wrong, i.e. poor performance [operations] Tempo, how can I change the scheduled maintenance?” he said. “How does maintenance he reschedule if it’s in a hot or cold or corrosive environment? That’s probably where we need to think.”
In the digital world, the Army “needs to learn how to use data and how to use data differently,” he added. “I admit, we still don’t fully understand [predictive logistics] I’m still out. I really need everyone’s help to think through this issue. ”
McConville and Wormuth said future uses for AI will also include talent management and recruitment. “There may be ways that artificial intelligence can identify high-quality leads and prospects in ways that humans can’t,” said Worms.
But McConville stressed the importance of having a “stakeholder” when working with AI.
“It may not be that person who actually does all the work, but artificial intelligence will help us do our jobs better,” he said. “But at the same time, we want someone who says, ‘Launch this weapon system,’ or at least thinks about it.”
Barrett echoed McConville’s remarks, saying: [AI] as a machine. But… guess what, ChatGPT tinkerers, it’s the people that feed the machine. ”
As the Army deploys AI systems, Easley said there are four tasks soldiers can perform to ensure the technology matures properly. Data collection and annotation. Use that data to train his AI model. Use your model to see if your model is effective. Helps improve the model.
The service does a “great job” in terms of data collection, but “the Army still has a lot of data that it hasn’t fully captured, which it can use to train its own large-scale language models,” he said. “To be effective in our domain, these models need to be trained. [them] Based on our data. Therefore, you should consider: What is our HR data? What is our healthcare data? What is our operational data? What is our intelligence data? How can we use these in a controlled environment to create better models?”
And the model needs to be rapidly trained and retrained on Army data to improve, he said. Using the example of a mobile phone’s restaurant recommendation algorithm, he said, “The reason this is so good is because he’s been telling me which restaurants I like around the world for 10 years.”
AI-enabled recommendations may be provided in the future, but weapon systems will always have human control, but “other systems that are less important… [the] Machines can make the decisions,” Easley said. However, he added, it will be up to humans to train the AI to a level where it can be trusted for Army missions. “I never question a mapping algorithm that tells me where to go in the city because I know it has better information than I do,” he said. But “what went into the model, how was the model trained, and how do you use it, how do you get the truth behind the data? And that’s all… I think it’s a human effort.” ND
topic: information technology
