Earlier this year, Defence signed an agreement with Deakin University for the development of artificial intelligence (AI) capabilities to “process noisy and dynamic data into information that will give military decision makers an edge”.
What data? Defence said only that this data could be multi-source, multi-modal, irregularly timed and spanning a prolonged period.
That could be satellite or aerial imagery, signal intercepts, cyber activity, or system sensor data from aircraft or ocean observations, all of which challenge the ability of human analysts to derive useful information in any reasonable timeframe.
Examining vast amounts of surveillance information has long been recognised as an area where AI can excel, producing useful conclusions and even actionable intelligence in real-time or near enough.
Deakin University’s Applied Artificial Intelligence Institute will apply next-generation machine learning to develop suitable models for this task.
There’s more. In the same announcement, Defence said South Australian University would develop a statistical machine learning algorithm using data from devices like smart watches to detect early signs of infection.
Future practical use might include early detection of chemical or biological threats. The obvious practical application is for deployed Australian Defence Force (ADF) personnel.
Both projects are part of the Defence Artificial Intelligence Research Network (DAIRNet) Phase II call out, with combined $1.746 million funding from the Next Generation Technologies Fund (NGTF).
This underscores that Defence is very seriously examining AI, with a diverse range of projects and substantial work internally through the Defence Science and Technology Group (DST), Office of National Intelligence, Australian Signals Directorate and others, in collaboration with universities and research institutions.
Then there are foreign partners. Though the AUKUS agreement will deliver nuclear-powered submarines to Australia, it’s also about technical cooperation in a range of areas, including AI, through what’s termed AUKUS Pillar II.
In their joint statement after meeting in Washington, US president Joe Biden and Australia’s prime minister Anthony Albanese reported ongoing progress across current capability development programs “including holding the first demonstration of AUKUS artificial intelligence and autonomous capabilities.”
The AUKUS fact sheet cites early work on accelerating adoption and improving the resilience of autonomous and AI-enabled systems in contested environments.
The prime example would appear to be the Boeing MQ-28A Ghost Bat, an advanced unmanned aircraft that could conceivably enter service in Australia ahead of the US.
Navy has a couple of programs for autonomous AI-enabled capabilities, the Ocius Bluebottle, a small sun, wind and wave-powered surface surveillance vessel, and the Anduril Ghost Shark, a robotic undersea warfare vehicle.
Army envisages a future battlefield in which autonomous vehicles play a role. It has experimented with M113 AS4 tracked armoured personnel carriers converted to optionally crewed configuration, featuring BAE Systems’ AI-enabled Vehicle Management System, a relative of the autonomy engine on Ghost Bat.
That’s some of the cool stuff. Much of Defence’s AI effort is directed to less tangible but still vital capabilities.
Defence’s AI for decision-making initiative aims to help commanders make the best calls in uncertain situations when data is limited. DST Group plans to fund up to 50 pilot proposals.
Respondents were invited to propose projects addressing one of 30 nominated challenges. Example: Using AI with unmanned aerial systems to locate victims of a mass disaster event.
The Defence Strategic Review, released in April, says demand for Defence’s capabilities innovation system has never been higher and relevant programs need to accelerate research and innovation relating to capability.
That’s consistent a consistent DSR theme – Defence needs to move faster in capabilities development and acquisition.
Four days after the DSR was launched, the government launched the Advanced Strategic Capabilities Accelerator (ASCA) with funding of $3.4 billion over the next decade. Obviously, not all is for AI research but it will absorb a significant portion.
Australian AI capabilities are significant and growing. Many civil AI applications are equally useful for defence and vice versa.
Defence’s DAIRNet was founded in 2021 as an initiative of the Department of Defence managed in partnership with the University of South Australia. It’s been active, listing 58 research projects, 11 current, with 71 organisations and 16 researchers engaged.
Their website says robotics, autonomous systems and AI form one of Defence’s Sovereign Industrial Capability Priorities.
DAIRNet says the ability of the Australian AI sector, including researchers and industry, to design and deliver these systems will enhance defence’s combat and training capability by improving efficiency, reducing physical and cognitive load to the operator, increasing mass, achieving decision-making superiority and decreasing risk to personnel.
None of that is revolutionary. For example, various drone systems do much of this already.
Analyst Dr Peter Layton says much of the ADF mainly saw AI as being used in human-machine teams to improve efficiency, to allow the force structure to be affordably expanded by adding many small AI-enabled systems, to achieve decision superiority by making better decision and to decrease risk to personnel.
“Importantly these uses suggest AI is understood as a technology to do things better, not to do better things,” he says in a paper titled Evolution not Revolution – Australia’s Defence AI Pathway.
China is regarded as a world leader in AI, though we don’t know for sure. According to Taiwanese tech executive Kai-Fu Lee, China developed an early AI edge thanks to decades of accumulating data from snooping on its citizens.
The US is a significant player – many will have tried out the ChatGPT program for themselves. This was developed by US AI company OpenAI. ChatGPT stands for Chat Generative Pre-trained Transformer.
