Op-Ed: AI moves into defense in Australia – The piece of string just got a lot longer

Australia’s new National Defence Strategy focuses on making any attack against Australia’s interests prohibitively expensive and risky – Copyright AFP Simon Wohlfahrt

The Australian government is looking at creating an AI-based super-cloud with Amazon Web Services for defense. The main subjects under discussion are money, capability, and role structuring for what is essentially a different world.

Excuse a bit of necessary preamble here:

Military AI is inevitable. Modern military and intelligence generate data on a truly colossal and ever-increasing scale. Real-time data needs are exponentially expanding as seen in the Ukraine war.

There is also a long ongoing and likely to be eternal narrative about AI in defense roles. Simultaneously there are also a lot of warnings coming from high tech corporate levels about the potentially catastrophic risks of AI to businesses and investors.

These warnings are not “ornamental”. The risks are real enough, and the last two years of constant warnings about the irrational hype about this very primitive pre-AGI level AI is well inside the ballpark.

Say you were setting up an AI cloud for your defense services. What would you need to consider? Say also that you’re finally at the completion stage of a long-running tortuous epic of overdue upgrades in capabilities.

A whole new tier of data management requirements has meanwhile just landed on your head.

That’s the Australian situation, and most other militaries are in much the same boat. It’s a technological minefield and a fairly apt metaphor. Will this tech work? Can it do what we need it to do in real-time? Can we trust it? What about security? There’s a whole rhetorical opera in there somewhere.

…Except military roles and operations are less based on aesthetics and press releases than operas. Under all these ponderously obvious questions is a real need for trustworthy answers.

That’s a problem.  Military tech tends to balloon and encompass “disparate” technologies almost daily. Data management in particular is downright brutal, and the security of these tides of information is critical.

As you can see, even in a couple of paragraphs; you’re talking about a pretty evasive piece of string that actually dictates even the discussion of military needs.  That’s likely to be an expensive problem.

The Australian Defence Force is what you’d call a core military of air land and sea capabilities. It’s combat-efficient and professional. It’s not a particularly “neurotic” organization. The culture is often criticized, but it works well on the job.

It is now about to take on a large array of new tech which will require murderous levels of testing by frontline people. Operational testing is traditionally a no-bull-tolerated exercise in the ADF, with good reason. Grunts get a word in edgewise.

Now, contrast the hype with the real needs. Current-generation AI isn’t at all impressive in so many ways. It can do some things well, mainly the decades-old stuff that’s been grafted on, like essay writing and deepfakes. It can definitely handle mega-data streams. That’s largely thanks to the many barely mentioned scientific AI applications which have also been around for decades in one form or another.

But, and it’s a huge but – Can it do the military roles on the many different scales required? In theory, yes. Can it handle multiple land engagements in real time? Can it manage drone fleets and anti-drone warfare? What about doing anti-submarine warfare, while looking out for the odd wandering nuke?

We now finally get to some fundamentals:

The capacity of defense AI will need to be peer-competitive. That means constant upgrades and counters to opposing AI. This is also unavoidable.

Personnel training requirements will also need to be ongoing and “exhaustive”. That’s in the sense of training people to interact with AI on a reflex basis. It’s an odd environment, so orientation to systems and new tech are built into that scenario.

There can be no set price tag. Budget estimates will have to be broadminded forever more. It’s a commitment to essential operational capacity. Australia’s relatively modest $10 billion current model is a realistic prototype of future iterations. The only real way around it is scheduled outlays.

There’s a lot for the new ADF AI to do already. Our region is pretty tricky at the moment. AI data management will be handling a mass of disparate data at all levels. AI could streamline this bulky, somewhat thankless issue very effectively.

The opportunities for monumental stuff-ups are legendary at this level. All new tech has teething issues, but on an overarching platform, they’re worse. Murphy’s Law is the only law never to be repealed, and that’s the big lookout.

If you remember a few years ago the big hoo-ha about “no robots in warfare”, you can see how that panned out. It is now setting up. With any luck, human intelligence will stay ahead.

_________________________________________________________

Disclaimer
The opinions expressed in this Op-Ed are those of the author. They do not purport to reflect the opinions or views of the Digital Journal or its members.