Co-authored by: Davidson President, MG John W. Holly (USA Ret.)
& Executive Vice President/Chief Technology Officer, Lisa Hammitt
The 2020’s will be the decade of artificial intelligence-driven innovation for the US Military. AI will not replace humans, but it will assist us in arriving at better decisions and complement the human decision-making logic.
AI’s proliferation is driven by a rapid paradigm shift from confrontation-through-force to leveraging intelligence toward appropriate responses; this trend maximizes gains and minimizes conflicts. The lines between espionage, intellectual property theft and national security conflict have all but been erased. Therefore, we need to further eliminate any barriers to innovation among the private and public sectors. Advancements in compute power, chip design and distributed networking capabilities have enabled widespread use of artificial intelligence in the private sector which now mean effective national defense is access to a level playing field.
Though a catalyst to the paradigm shift, AI is not a panacea. It is best characterized as an enabler which, when combined with other disruptive technologies, drives ‘hunt forward’ missions.
Growing constraints on defense budgets, cloudier definitions of digital sovereignty and mounting requests from allies, both traditional and new, to hunt on their networks amounts to sharpening the focus on AI.
What do we need to consider as we confront these myriad challenges? Are the risks adequately addressed? Do we have the appropriate AI capacity? Have we trained and tooled our personnel to be data-driven? Are we optimized to the best private-public partnerships?
If the prominence of Cyber Space continues to shape defense policy, are we in a Boolean where it is an AI-driven mission that indicates whether we are at war or not? And, unlike traditional missions, can AI really give us an equivalent to a maritime experience?
These topics showcase how our military is modernizing itself with AI; however, the same distinctive components still underpin a traditional mission: clearing a hostile building and providing repair parts to tactical units, to name a few, and we are better served to think of how AI can enable them.
General Principles
Let’s look at what powers AI. Data are everywhere. But how does one mine a vast array of data inputs available to military operators to deliver them across a military enterprise? By transforming them into insights that drive appropriate action to affect a desired outcome.
A lofty goal, to be sure; AI differs from the deterministic logic of its software predecessors by ingesting large data sets to produce metrics. It augments human reasoning in tasks that are too costly to be carried out by hand. What it is not a replacement for is human judgement, particular that of a highly-skilled, mature operator executing a mission.
Small But Sufficient Data Sets:
Advanced data analytics differ from Artificial Intelligence/Machine Learning. Data Analytics takes input from disparate data sources and condenses them into more intuitive output, such as key metrics on a dashboard. Conversely, AI takes data and force-ranks recommendations that predict the best outcome for a given situation, and potentially, generates both cause analysis along with courses of action – it yields insights through data.
Though the current focus is on Big Data, or, on computing massive amounts of data, the greatest gains in military use cases are made in augmenting human cognitive capabilities with discrete and targeted data sets. It’s all about the data and, with 20 billion devices coming online through the Internet of Things (IoT), edge computing using smaller data sets becomes the new archetype. While AI-powered algorithms make continued progress with the addition of more data, speed to execution in the mid-2020’s will dictate success.
Integrated AI & Cyber Framework:
AI is not a standalone capability. To fully exploit the inherent potential of AI it must be connected into architectures, or substrates of multiple networks, that inter-communicate among disparate inputs of data, all revolving around a tactical engagement and the status of its logistics support.
For AI to operate at scale, it needs to be part of an integrated cyber security framework where data protection is integral to the network’s operations. As threats surface from many different sources, robust AI algorithms, along with the data sets feeding them, must be rendered impermeable to either directed or accidental cyber threats. A secure ‘mission accomplished’ then translates to innovations in encryption, masking, key generation and distribution that take users to data.
Answers vs. Decision Aids
Integrating AI into the cybersecurity network provides benefits beyond security. One current drawback of AI in centralized systems is the output usually provides a rear-mirror view of events, only after consolidation of the data into a structured format. Most Warfighters require both real-time and continuous analysis that aid in making decisions under highly uncertain conditions. Data on real-time alerting and monitoring dashboards are frequently and predominantly populated by unstructured data where analysis is in-stream.
These advancements have placed answers at a Warfighter’s fingertips, but in missions where the impact of risk-taking is high, any human being needs a high degree of trust and transparency around the algorithms output. Therefore, real-time analytics in warfighting scenarios without cause analysis or explanations of how AI force-ranked its recommendations are unlikely to be adopted.
Like many domains across the warfighting enterprise, 97% of data are dark and unused, but low-hanging fruit remains. The administrative enterprise or back office can optimize the logistics of any operation where AI is a service and simultaneously modernizes antiquated and brittle infrastructure into a streamlined, mission-by-design architecture.
Tactical Timing
In a tactical situation, accurate, understandable, real-time information is the key to success. Fed by sensors, detectors and network platforms, we are in the age of plethora. Since every device has the potential to speak to the Warfighter, how do we make it meaningful? Moreover, how do we phenomenally mine the data so that we surface the insights that are most relevant to the decision at hand?
Latency is paramount. If any system is forced to sift through all possible search paths, it will not be timely enough for a Warfighter’s need to act. In planning and optimization algorithms, AI techniques allowing an autonomous vehicle or an automated chess bot to make moves in split-second response were invented more than 50 years ago. The underlying enabler to bringing these techniques to the Warfighter will hinge upon whether we can push the sub-routines into the actual silicon. Edge computing and 5G allow for autonomous AI chips containing core learning libraries where multiple nodes distribute moves across intelligent agents. These agents act with surgical precision in looking only for a specific goal while ignoring all fruitless search paths.
In a tactical situation, analyzing every single available piece of data is not realistic or rapid – near instantaneous responses are required. The best AI can hope for is to be a digital concierge to the Warfighter where the most important moves are put forth in a natural manner.
Specifically, if 5G can handle up to 1 million connections in one square kilometer, AI would help us redraw the chess board simply by capturing an adversary’s 5G network. More discretely, if we began to use AI-enabled Cobots (robots designed to specifically work in conjunction with humans) we knock down doors, then locate and move to the best supporting positions so as not to inhibit tactical movements of human operators. In both cases, a human responds with surgical precision to the best opportunities before them, whether the implications be large or small.
Adaptation and nimbleness are crucial. As large, cumbersome organizations transition to smaller entities which are better able to pivot, agile principles and practices will follow. Weaving AI into the mission can be a catalyst to architecting for both flexibility and innovation.
Strategic Decision Support
The private sector has reached Moore’s Law and, in strategic military settings, it is no different. Data is exploding and we haven’t yet harnessed the power of AI to manage it. As our operations become increasingly complex and multi-disciplinarian, are there any silver bullets?
Innovation Through Data
Just as AI can be a digital concierge for our Warfighter, it can equally help us to map out the thought process and strategy of the enemy. Our adversaries don’t always think the same way we do, nor do they arrive at conclusions we might view as acceptable. Eliminating our internal biases in a strategic environment can be a dramatic force multiplier. We may not always understand how AI reaches a conclusion. Our logic is guided by our interpretation of our experiences. Similarly, AI merely takes the data in an unbiased manner and learns how an intelligent adversary processes the information to reach a conclusion within their frame of reference. Real-time modelling and simulation of enemy actions can prove to be an innovation that is exploited to our advantage.
Hyper-converged Connectivity Instead of Data Silos
We must steer clear of bespoke solutions uniquely developed for a military application. Humans all have brains which operate in a similar manner. We develop expertise based on our experiences. We must think of AI as analogous to a human brain, learning through the experience of rapidly sorting through vast amounts of data.
Summary
AI is the catalyst to Reshape the military workforce and warfighting force. However, we must recognize that today’s applications are equivalent to the mental capacity of a 2-year-old. How, then, do we begin to shape AI into a teenager? What do enabling technologies such as 6G hold in store? 5G is a near-term disruptor, but where should our investment priorities lie to readily exploit other emerging technologies as they mature?
This decade will develop a codified culture of AI exploitation and use. We shouldn’t view AI through the lens of a division of responsibilities between humans and machines. We should instead ask ourselves how the military benefits by a complementary balancing of AI and the soldier. So, is AI a silver bullet? Not quite, but possibly a copper or bronze one.
コメント