1) ARTIFICIAL INTELLIGENCE
With the possible exception of big data analytics, no other EDT has more cross-cutting implications for military operations than AI, which derives from the ability of algorithms to make optimal or quasi-optimal choices to achieve specific
goals.
Combining AI with other technologies and functions will yield new military capabilities that were previously found only in the realm of fiction. Using it to crunch big data, for example, will enable fast decision
making and animate sensors, the testing and application of new materials, or the rapid-fire coordination of fleets of autonomous military platforms.
Applying AI to logistics is already generating significant improvements to operational efficiency and military supply lines, and thus helping reduce costs. Building AI into a soldier’s
suite of sensors, for example, would yield more effective reaction time in the field while enhancing communication and data flows at all levels of the chain of command.
AI will inevitably shift into predictive battlefield assessments while vastly accelerating real-time situational awareness in all operational domains. That will add up to faster and better-informed decisions for military
and political leaders. Meanwhile, future progress in machine-learning and quantum computing should further advance the speed and efficiency of nearly all military tasks, making AI the most versatile EDT.
How EDA contributes
EDA is laying the groundwork for Europe’s armies to exploit AI in many operational areas. Based on an EDA action plan, approved in December 2020, three major projects will be launched in the coming months.
The first project will explore the concept and rules for an EU-wide pool of defence data, guided by the principles of sovereignty over data, security and trust, data interoperability, and the portability of data and services.
“The idea is that each Member States and its industry could contribute data that is anonymised and then made available to all others for purposes of research and simulation. Here we’re mainly talking about the development and testing of new algorithms”, said Kikiras,
adding that the Agency will soon send a request for information to the industry, research, and defence ministry stakeholders for feedback on the four principles. Their views will be collected and reviewed by the
end of 2022, after which the Member States will decide on the way forward.
The second project will focus on analysing the requirements for defence-trusted AI. This pertains to matters of human oversight (known as the human-in-the-loop rule), the technical robustness and safety of operational AI,
traceability and accountability, and the overall rules of data governance.
The third project aims to map out the requirements for a unified EU framework to validate and certify military AI-based systems. Harmonised certification would help confront threats in areas such as cyber security
where fast-tracked solutions could be fielded more quickly such as enabling networks to self-configure/self-patch when detecting vulnerabilities.
By its very nature, however, this project presents daunting challenges since any shift toward pan-EU certification – whatever the sector – cuts across huge numbers of institutional, regulatory,
and industrial players. Fortunately, there are precedents to emulate.
“We will turn to the European Union Aviation Safety Agency (EASA) and the European Union Agency for Cybersecurity (ENISA) to see how they developed their methodology and principles of validating and certifying AI and then apply them as appropriate to military AI,” observed Kikiras. “It’s not just about the process of validation, but infrastructure as well – the facilities and how you test against reference AI models. For example, when testing an intelligence-gathering swarm of drones, what reference models to use? These are some of the issues that need resolving.”
One encouraging development is a recently completed two-year study by EDA and the European Space Agency (ESA) known as “Cyber-Defence for Space” that analysed the cyber threats to ground
and space infrastructure. Its results were presented in early November, along with recommendations for creating cyber-operational centres services for cyber threat intelligence for space, utilizing AI technologies.