The Army’s 100-day plan to find barriers to AI adoption

Poisoned data sets, Trojans and ever-evolving cyber threats could derail the military’s plan to adopt AI broadly. But a new 100-day plan aims to remove obstacles and prepare the service to scale third-party models and algorithms.

The plan, released in March by the assistant secretary of the Army for Acquisitions, Logistics and Technology, comes after the Army released a software guidance that aims to help the service – and industry partners – deliver capabilities more quickly by following to mimic how commercial developers work.

“The Army recognizes that we are not going to do the development and training of algorithms and models better than the industry, we want the industry to do that. And they do it extremely well. And so we want to adopt a lot of that. And one of the obstacles to adoption is: how do we look at the risks surrounding AI?” Young Bang, the Army’s principal assistant secretary for acquisition, told reporters Monday. “That means: what are the issues around poisoned data sets, hostile attacks, Trojan horses… And it’s easier to do if you’ve developed it in a controlled, trusted environment that is, for example, owned by the Defense Department or the military.”

These efforts will ultimately align with the military’s ongoing efforts to simplify the way it uses AI, particularly for intelligence systems, through a program called Project Linchpin.

“There are a lot of programs right now working on models, training, deploying and testing them” as the Army develops the Linchpin environment, Bang said.

Linchpin is designed as a digital environment with vetted tools, infrastructure, standards, possible use cases and associated data for AI, said Bharat Patel, head of the Army’s Project Linchpin and sensor AI program. And having all that in one place should make it easier for project managers to integrate artificial intelligence into programs that can benefit from it.

Right now, Linchpin is focused on defining those use cases and preparing the associated data, says Patel, who is also product lead for the Intelligence, Electronic Warfare and Sensors Program Executive Office.

“We do basic model training. But it’s not for performance or anything. We’re trying to figure out what our process is, what our governance is, what our standards are that will allow us to do this much faster. So our goal really is to learn something and make sure that once the 500-day implementation plan, the contracts and everything that’s in place, we’re ready to go,” Patel said.

And as Linchpin matures, the plan is to add algorithms from other programs and third parties so program managers can pick and build on capabilities from there. Bang said details about Linchpin’s readiness will follow after the 100-day plan, which will likely be completed this summer – with a 500-day plan to follow.

“Every theater is different,” Patel said. “You can’t think that a model for (European Command) out of the box will work for (Indo-Pacific Command). The trees are different, the biosphere is different… that’s why it’s super important to get behind the use case. And where that (area of ​​responsibility) is specifically located. So we are looking at that very closely.”