Inside America’s plans for an autonomous, AI-powered military

The United States is accelerating an effort to revolutionize modern warfare by fielding swarms of self-operating drones and weapons systems. The push will shape the next generation of war and, military leaders hope, give America a leg up on China in the new global arms race. 

With the Pentagon’s new Replicator initiative, the U.S. is moving fast toward an ambitious goal: propping up a fleet of legacy ships, aircraft and vehicles with the support of weapons powered by artificial intelligence (AI), creating a first-of-its-kind class of war technology. It’s also spurring a huge boost across the defense industry, which is tasked with developing and manufacturing the systems.

For those watching the U.S. defense and security field closely, Replicator was greeted with a sigh of relief — though it’s also raising a host of concerns related to accountability and the human cost of autonomous warfare. 

“This is the same as the transition from crossbows to guns, from cavalry to tanks,” said Steve Blank, co-founder of the Gordian Knot Center for National Security Innovation at Stanford University. “Thousands of these things that are semi-autonomous or autonomous is a major transformation in warfare. Period. All nations will eventually get here.”

The U.S. is hoping that Replicator, designed to field thousands of fully autonomous systems within two years, will get the military ahead of foreign adversaries who are also pursuing this technology, particularly China. 

In the event of a clash over the self-governing island nation of Taiwan, AI-powered drones and aircraft will be vital to countering a larger Chinese military force, especially over contested waters and airspace where Beijing’s mass could overwhelm American forces.

Moreover, the war in Ukraine has shown just how important drones are. While those deployed in Ukraine and Russia are being guided remotely by human operators, AI-powered ones would significantly boost the capabilities of the forces deploying them by swarming targets.

Replicator also means faster and cheaper military technology. It’s spurring a huge boost across the defense industry, giving smaller defense tech companies a chance to compete for contracts against traditional defense giants.

If successful, the U.S. will be able to quickly produce these drones for a much cheaper cost than conventional systems, putting the military at a competitive advantage against other rivals.

The only question for defense contractors and war analysts is whether the U.S. can meet the program’s ambitious timeline.

Deputy Defense Secretary Kathleen Hicks announced the new initiative in late August and said the goal was to develop these weapon systems in the next 18-24 months. She added this month that Replicator does not involve new funding or personnel, but rather a steering committee will guide its progress across the entire military.

Eric Pahon, a spokesperson for the Defense Department, said it was mostly a matter of scaling out technology. He admitted there were some challenges with bureaucracy, but said the U.S. has a history of being able to pivot.

“A lot of this is about cutting the red tape within the department,” Pahon said. “We have the ability to do these things, but it’s going to take a massive mindset change and somebody who’s able to help kind of cut through the biggest delays that we have in the department.”

Pahon said Replicator would put the U.S. ahead of China and allow for AI weapons systems to be mass-deployed in nearly every field of warfare.

“The great thing about these kinds of cheap systems is that we can really easily reprogram them and refit and be really creative,” he said. “I can send an [autonomous] drone to carry a bomb or it can do weather readings or it can do surveillance. I’ve got thousands of them, no big deal.”

One challenge might be getting all of the technology to talk to each other, Pahon said, but there are plans to invest more heavily in Joint All-Domain Command and Control, a computerized network in the early stages of development.

Over time, Congress will need to step in with more robust funding for the next-generation weaponry and support structure. It’s not clear if AI weapons will end up as an immediate priority in the defense spending bills currently caught up in budget negotiations on Capitol Hill. 

Last year’s defense bill included a requirement to enhance warfighting capabilities with AI and report annually through 2025 on progress.

Replicator inherits the original goal of Assault Breaker, a U.S. military program in the 1970s that proposed methods of using swarms of weapons, like long-range smart weapons that divide into tracking submunitions, to counter a larger Soviet Union until reinforcements could arrive.

Assault Breaker 2, similar to its predecessor, appears to have been rolled out in the last decade. Testimony from Steven Walker, then a leader of the Defense Advanced Research Projects Agency, in 2019 said the Pentagon was focused on using “emerging technologies” in the strategic initiative.

Recent advancements with machine learning algorithms and cloud technology has expanded what’s possible. AI weapons have become a dominant research topic for the military over fears of China overpowering the U.S., particularly in the maritime domain, after Beijing’s Navy became the largest in the world.

The naval concern sparked the idea that Washington should focus more on smaller tactical units rather than following the tradition of building more large naval ships that are vulnerable to attack in a Taiwan scenario.

The Navy has already reached operational capacity with fully autonomous watercraft with Task Force 59. In the successful exercises, the Navy has worked with AI systems that carry cameras and transmit images to the fleet.

The Army is also pursuing ground-based AI systems, and the Air Force has been experimenting with Collaborative Combat Aircraft, a program designed to deploy teams of autonomous systems to work with human-operated fighter jets, which supporters argue will completely change the nature of dogfighting.

The Air Force is asking for $6 billion over the next five years for the program, according to Air and Space Forces Magazine, and plans to integrate the systems into its arsenal by the end of the decade.

A leading company in the development of this technology is Shield AI, which in 2018 fielded an AI-driven quadcopter, Nova, to help scan and identify combat areas.

Shield AI employs software technology called the Hivemind, which it claims as the first and only AI pilot system to ever be deployed in combat.

Hivemind has even operated an F-16 and could feasibly power ships, satellites or submarines. In August, Shield AI completed a successful exercise of autonomous craft teaming up through software to accomplish a mission, much like they would in combat.

Brandon Tseng, co-founder of Shield AI, believes the technology is “mature” enough to make Replicator happen. 

“It’s about allocating the resources under the right capabilities and products to drive scale,” he said.

Replicator promises to be a boon to smaller defense tech companies like Shield AI, which are outpacing traditional defense contractors in the AI field. 

“Their business models are built to build those exquisite aircraft,” Tseng said of the industry giants. “What you’re talking about is a fundamental shift in the defense industrial base and the operating paradigm toward cheaper, more affordable aircraft.”

Raytheon, Northrop Grumman, General Dynamics and Boeing did not respond to requests for comment on this story. Both Lockheed Martin and General Atomics said they were in position to work with the Pentagon on Replicator.

Defense tech companies have long been calling for a focus on AI weapons and, though Replicator is being applauded, they are asking for the Pentagon to do even more as concerns grow about military vulnerability in a clash with China.

In a letter to President Biden this month, defense tech firms asked to codify and accelerate Replicator, modernize defense trade laws and regulations and step up investment in AI, among other urgent changes.

“Despite these important efforts, in too many cases, our fighting forces are equipped with tools decades behind commercial technologies most Americans use daily,” the letter reads. “Most new cars on the road today have more advanced AI than deployed U.S. military vehicles.”

On the defense tech side, companies are not just talk: They say they are ready to meet the demands of Replicator.

Skydio, a signatory of the Biden letter, uses AI for smart drones that can overpower traditional defense systems. Skydio’s president of global government, Mark Valentine, said Replicator will shield against electromagnetic defensive technology that can disable drones, mostly because the sheer size of swarms make those defenses moot.

Valentine said to counter growing threats in Beijing and Moscow, the U.S. “must build smart systems at extraordinary scale, and we must do it now.”

“We recently increased our manufacturing capacity by 10 times and are well prepared to answer Replicator’s call,” he added. 

Valentine also wrote a blog post earlier this month highlighting the crucial aspect of achieving cheap, affordable mass through autonomous systems — but also ensuring they had key capabilities such as target tracking and convoy support.

“The concept of ‘affordable mass’ has received a lot of attention as of late,” he wrote, but “smart drones are required for effective affordable mass.”

In Utah, Teal Drones is whirring a major factory to life to meet Replicator’s ambitious timeline, said George Matus, its founder and CEO.

“Our current facility can produce up to a thousand systems a month,” he said in a statement. “As we scale toward that number, we will be able to expand our facility, team and systems.” 

AeroVironment CEO Wahid Nawabi also told The Hill in a statement that his company will be able to meet demand and make autonomous drones “in large volumes at the levels of reliability” expected from the Pentagon.

But it’s not just drone-producing companies that will benefit and work with Replicator. The U.S. will also need strong defense against enemy autonomous armies. 

Epirus is a defense company that builds AI-powered systems to counter drones and other unmanned aircraft, using a high-power microwave beam emitter to disable swarms of drones.

Mara Motherway, the chief growth officer at Epirus, said the company was part of a “layered strategy” working in concert with offensive drones.

“We see it as complementary, and we’re excited about it,” she said. “As they start to produce those, [we ask if] there is something we might be able to help with that is matched to that aircraft.”

With Replicator, the Pentagon does not aim to phase out legacy aircraft, ships and vehicles, but will seek to complement them on the battlefield. The stakes are high. 

Blank, from Stanford, said if Replicator is realized, the U.S. could reverse the last decade of America’s military rivalry with China. 

“We will have done to their Navy what they just did to ours,” he said. “Now, all of a sudden, we made their service obsolete.”