U.S. ‘Not Ready’ for Rapid Battlefield Innovation, Industry Exec Says
FORT WORTH, Texas — The United States is not currently prepared to adopt the rapid innovation cycles Ukraine has showcased in its fight against Russia, but artificial intelligence presents opportunities for the Defense Department to greatly reduce its innovation timelines, an industry official said Oct. 1.
Will Roper, founder and CEO of software startup Istari Digital, said visiting the frontlines in Ukraine and seeing warfighters bring new capabilities to the battlefield in two-week cycles “immediately raises the alarm bells that we are not ready for this.”
Roper is the former Air Force assistant secretary for acquisition, technology and logistics and founder of the Defense Department's Strategic Capabilities Office. During his Pentagon tenure, he championed the adoption of digital engineering in the defense sector.
In particular, the use of uncrewed aerial systems as “bullets” and how the battlefield is being reshaped around drone warfare and “these fast cycle times, we’re not ready for that,” Roper said during a fireside chat during the National Defense Industrial Association’s Future Force Capabilities Conference and Exhibition.
“And if AI were also part of this — which it would be in a future war — the cycle times would go from two weeks, which are the human limit, to something much faster,” he said. “If you step back from the U.S. military, or from NATO or our allies and partners in the Pacific, we don't have the infrastructure to even begin attempting that.”
The United States and its allies need to prove out these rapid innovation cycles “outside of laboratory Petri dishes, where we can prove something's possible but not scale it,” Roper said, and the “best way to get ready is to be your own adversary.”
“Make them as tall as you think battlefield needs require,” he continued. “Train against yourself and prove you can do the cycle time in realistic conditions” and then continue to bring the “timeline down until you feel like it's something you could win with.”
Artificial intelligence will be key to making these rapid development cycles a reality, Roper said, but the U.S. military must also be prepared that AI models will likely fail on day one of a future conflict, “because every side will be able to create conditions that weren't part of the training set.”
“Both sides are going to fail miserably,” and then they will collect data, retrain the models and push out an update, he said. “The side that can do that the fast[est] will go from being like an algorithmic dud to algorithmically dominant.”
While that playbook is “uncertain … the only way to get ready for it is to do it,” he said. “That's what I would wish, is that we were out operationalizing AI in real hardware, in real red and blue exercises and starting to train and learn how to hopefully predict these algorithmic windows of advantage if that's possible — and if it's not, then being the most agile to take advantage of them when they exist, repeat, repeat, repeat until the conflict is over, and [that’s] just not the playbook today.”
One challenge is that while AI adoption is quite high, there is still a significant lack of trust in the outputs of the models, Roper said.
Having a digital infrastructure that provides confidence in the AI models to both the government and industry partners in the development cycle would be a major breakthrough, he said, adding that Istari believes it has a system that could solve this challenge.
“We've got companies that are here today” — who Roper declined to identify — “that are using it at a significant scale — and I'll have more that will come out this year on the breakthroughs we've made — but it was far, far more challenging than I would have anticipated to take what works centrally and make it work de-centrally, so that … everyone can say that they were accountable for their data, their authority, but still have it interact with others in a way that both sides can independently verify,” he said.
Closing the “air gap” between warfighters and their industry suppliers — as Ukraine has done — will be key to any future conflict, Roper said.
“If a future war happened, I expect we would see a lot of the things we see in Ukraine today. We would see technology companies and operators working together as if they're one team doing DevOps — but not just for software, for hardware as well,” he said. “The air gap would have to be completely eliminated because the problems are moving that fast.”
And to truly move at the pace of innovation, “we have to go … past the human limit, because we are letting things go automatically and doing that in a way that — it's not risk-free — but it's at least accountable and responsible,” he said.
“There's just a whole new playbook there. It will be an uncomfortable playbook for many, because we'll have humans out of decisions we normally would have had them in, but if they are in, we'll lose,” he said.
While AI is not perfect, it is “going to keep getting better,” he said, and the United States owes its warfighters the capabilities that give them the “most options possible. And those options need to be not just stuff, it needs to be time as well.”