OTTAWA — When Ron Lloyd thinks about how the Canadian military could be using big-data analytics, he imagines a young sailor nearing the end of a gruelling series of on-and-off watches in the middle of the ocean.

“You’ve been at sea for 30 days, you got about a three-metre, four-metre sea, and it’s in the middle of the night, you’re probably not operating at your best,” the retired vice-admiral said in an interview with The Logic. “But if you’ve got machine learning or an algorithm that can help point that sailor to something that they might have missed that would be anomalous, then that’s a really good use.”

Until 2019, Lloyd was commander of the Royal Canadian Navy, and while holding that post he did a stint as the Canadian Forces’ first chief data officer.

“When it comes to supply chain, people management, asset management and capital asset management, we’re very, very much the same as industry,” he said. “The only difference is at the end, in the battle spaces, where we’re fundamentally different.”

Sarah Shoker, who researches the use of artificial intelligence in defence at the University of Waterloo, said the quantity of data generated by a modern military is already surpassing the capability of humans to examine it.

“If you look at the U.S. case, for example, the vast majority of the data that’s collected by fighter drones is unanalyzed, because it’s impossible, using any kind of human physiology, to analyze the amount of imagery that’s collected by a drone in a particular year,” she said.

That’s just one function. David Perry, who specializes in defence procurement as a senior analyst at the Canadian Global Affairs Institute, called the Defence Department “a 100,000-person workforce that does everything from run fighter jets that have to order ammunition to run schools and police systems.”

It tracks weather, examines water columns in the sea, maintains buildings, trains recruits, checks propellers for wear, buys uniforms, surveils potential threats. Sometimes it even fights.

Sifting all the information it generates demands so much computing power that Lloyd said even the Department of National Defence cannot practically sustain it.

“When large software providers can’t compete with the Amazons and the Googles and the like, in terms of cloud services, what would make us think that in the military or government, we would be able to compete, or should even be operating in that space?” he said.

All this points toward a future where basic elements of our national defence rely on computers not just supplied by foreign-based companies, but operated by them, as well.

The Department of National Defence declined an interview request on the subject, but said in a written statement it currently has both public and private clouds, for uses ranging from Microsoft apps for routine office work to custom-made secure military functions.

“DND/CAF sees great potential in emerging information technologies such as cloud to meet its various mandates and protect Canadians. Cloud computing offers key benefits in terms of IT efficiency such as the internet of Things (IoT), expanding Edge computing capabilities, and future access to big data analytics, machine learning, and artificial intelligence to enhance decision-making,” said communications officer Shania Mahendran.  

Amazon Web Services, the company’s cloud-computing arm, apparently sees an opportunity in Canada, recently posting an opening for someone to promote its offerings to Canadian defence and national-security agencies and try to influence policy and regulation its way.

The company declined to talk about its intentions, but it has aggressively pursued similar business in the United States.

In 2019, the Pentagon awarded a 10-year contract to AWS rival Microsoft for what it called its Joint Enterprise Defense Infrastructure (yes, “JEDI”) project, an effort to create a supercloud for the American military. Amazon went to court to complain that then-president Donald Trump’s hate for its founder Jeff Bezos was an unfair factor in the decision, and in early July, the Defense Department blew the US$10-billion procurement up.

Blaming “evolving requirements and industry advances,” the Pentagon said the JEDI plan would no longer suit its needs and will be replaced by the “Joint Warfighter Cloud Capability,” with components awarded to both Microsoft and Amazon. A pre-procurement notice says they’re the only two “hyperscale” cloud providers based in the United States with the capabilities the U.S. military needs.

In Canada, AWS has had a contract since 2019 to provide the government with cloud-computing services for data with “Protected B” status. That’s not a national-security classification—those secrets use a different scale—but it does allow Amazon’s servers to handle material that “if compromised, could cause serious injury to an individual, organization or government,” such as human-resources data and trade secrets, on server farms in the Montreal area.

The Canadian military might make a very appealing customer for much more.

Publicly available plans from the Canadian Forces are replete with ambitious vagueness. For instance, the third edition of the army’s Advancing with Purpose modernization strategy, released in early 2021, talks about making sure that “the One Army team, leaders, and, in particular the soldier, will have access to the data they need to inform decision-making to identify opportunities and improve operational effectiveness at the speed of relevance.”

But the army’s plan does include some short-term specifics. The Defence Department is getting ready to buy simulators to train soldiers for the varied duties on combat vehicles, and by 2026 the army hopes to use big-data analytics to tell when the trainees are ready to move on to real tanks and troop carriers.

The navy’s 2017–2022 strategic plan talks about using data to optimize fleet deployments and maintenance—mundane, maybe, but critical to keeping the fleet in good repair while also having the navy available to do things.

Lloyd imagines a ship’s systems detecting impending equipment failures before they happen, assisting sailors in digging the parts they need out of storage and guiding them through repairs with augmented reality, then ordering replacement inventory to the ship’s next port.

“Over time, if this was the fourth time that pump has failed, is it because of the pump or maybe the operator is doing something wrong? So you can go back through those files to see if it’s that operator with a problem, or maybe the plant maintenance routine was done poorly,” he said.

Then there’s the spear’s pointy end. We’re already in an arms race: Shoker said China sees military artificial intelligence as a potential strategic advantage.

“China’s perception is that they know that their military tech is very outdated. And so instead of trying to run and catch up with the U.S., they are going to use AI as a leapfrog technology to modernize, to catch up, instead of going through the traditional development route,” she said.

None of the people The Logic spoke to endorsed the idea of “killer robots,” weapons systems that can make their own decisions to shoot. (Canada is officially against them.) Just using algorithmic assistance for human decisions can be problematic: much of Shoker’s work is about guarding against biases that creep in from faulty inputs or ill-considered processing methods.

But the pileup of security-related data, however it’s used, is a real problem. 

A civilian car can generate gigabytes of data in a day. Now imagine a fighter jet loaded with cameras and sensors and weapons systems that can also deploy a cloud of drones, multiplying its own data-collecting capability. Or combat between semi-autonomous planes.

“Sort of like crazy science fiction coming to life. All of that is going to require a lot of compute power,” said Christopher Parsons, a senior researcher at the University of Toronto’s Citizen Lab, who focuses on technology and national security. It will take vastly more than any such craft could carry aboard or a battalion could haul around.

Not necessarily a constant amount, though.

“It’s really expensive for government procurement to go out, buy a whole bunch of stuff, it sits in a basement until it gets fired up, and it goes dark again, and it gets fired up again,” Parsons said.

“Alternatively, you don’t buy enough or you don’t buy fast enough. You’ve got all this data, and you’re collecting more data, but you don’t have any place to put it for the next two months, while you’re waiting for some hard drives or something like that to come through, or you don’t have enough compute power. Canadian Special Forces are deployed somewhere and you really need to be able to process the data they’re collecting in real time, and you can’t do it.”

Server farms run by third parties that offer storage and processing capacity on demand are an obvious response.

The U.S. move to private contractors for this gives Canada implicit permission to do the same, Perry said. To take proper advantage, though, Canada’s notoriously obtuse system for buying military gear will have to change.

“The way that we’ve tended to write these things in the past is like, somebody would sit down and, to oversimplify it, they’d write a bunch of specifications for what they want the software to do. And then it’s got to chug through the Government of Canada’s whole system. So four years later, you’re at the point where you’re putting out a tender for something that you wrote a requirement on four years ago,” Perry said.

Instead, he said, the government will have to pick a vendor or vendors, enter into a long-term relationship, then work out together what the provider can do to suit Canada’s needs and how to revise the arrangement as time goes on. As the U.S. is doing.

Getting the procurement right is just the start. The navy’s strategic plan warns plainly about ongoing security risks.

“Potential adversaries are also exploiting technology and in this light, interoperability and systems integration amongst our closest allies will be critical to force design,” it says. “Managing big data, limited bandwidth, and the cyber vulnerability that this entails, will demand thinking and approaches that break with traditional mindsets. In other cases, for example watchkeeping at sea, we must preserve the age-old practice of heads-up navigation, and safeguard ourselves from an exclusive reliance on sophisticated technology that can fail, be spoofed, or purposefully hacked.”

In 2018, the Communications Security Establishment said Canadians had been among the victims of a hack—which it attributed to the Chinese government—targeting unnamed managed-service providers, of which cloud-computing services are one type. Affected clients included a global financial firm and technology and manufacturing companies, according to an indictment filed by U.S. authorities.

Parsons said the United States requires providers of sensitive cloud services to put their servers on secure government property, behind government firewalls. That limits the savings from sharing the servers with non-government customers, but it’s safer, and the U.S. government still gets the benefit of the private sector’s technological advances.

But any system that has data moving in and out is not impregnable to hackers.

“Are they trying to get into the government of Canada’s system? Maybe,” he said. “But also, they could just be trying to breach Amazon’s network and either exfiltrate data that hasn’t been adequately secured, encrypt data [so its owners can’t use it] when there hasn’t been adequate virtual separation, or otherwise just cause holy hell, to be frank.”

Lloyd suggests dividing the material up.

“Anything we can get into the cloud, off-[premises], into a third-party provider would be probably the most effective and efficient way of going, recognizing that your more sensitive highly critical national-security-type information you’d want to have locked down, much like any industry would have,” he said.

However we do it, Lloyd said, data and processing it are already as important to militaries as leading-edge weapons and sensors used to be.

“Getting that right is going to be really, really important for [the] success of the organizations in the future.”