An A3IC Agency is key---leadership in development of artificial intelligence by setting guidelines and procedures ensures cohesive development across NATO
Gilli 20 [Andrea Gilli December 2020 “’NATO-Mation’: Strategies for Leading in the Age of Artificial Intelligence” NATO Defense College Research Paper No.15 pp. 35-40 https://www.ndc.nato.int/download/downloads.php?icode=671] -os-
Championing innovation: artificial intelligence, integration and the implementation-enabling centre (A3IC) While innovation is often treated and discussed as an outcome, it is in fact also a process whereby champions – whether individuals or organizations – promote and implement changes leading to performance, mission or operational improvement.116 Within NATO, many stakeholders have an inherent interest in AI and in promoting the AI agenda. However, there is no clear “champion” whose goal is to steer the Alliance’s approach to adopting the technology, promoting the necessary reform, and devising the best practices for its employment. For this reason, the process of “NATO-mation” could benefit significantly from a centre specifically intended to serve this goal – tentatively referred to here as an Artificial Intelligence, Integration and Implementation-Enabling Centre (A3IC). Such a centre could lead adoption of AI for the Alliance, support Allies in their own adoption strategies, and connect the relevant offices and institutions at both the national and NATO levels. With a focus on training and interoperability, an AI champion for NATO could ensure that the Alliance treats innovation as an ongoing process and disseminates successful outcomes. The innovation process often requires a person or an organization mentoring, supervising, advocating and protecting innovations and innovators. Having an innovation champion is important to promote those micro-changes that often permit the successful adoption of an innovation: this is achieved by aligning the interests of all the individuals involved with the overall goal, so that resistance, opposition and sometimes even boycotting are minimized and addressed.117 While the buzzword “innovation” is generally perceived favourably, resistance often emerges because of cultural barriers such as opposition to change, sociological factors such as group identity and concern for loss of status, organizational dynamics such as career advancement being hindered by a new technology, as well as for cognitive and psychological reasons.118 The old adage goes: the thing people really hate, more than the way things are, is change. Military innovation poses even more subtle challenges. Adoption of AI may be perceived as especially daunting, because the technology is intangible and difficult to quantify: we can neither touch it nor see it. Additionally, previous lessons of military innovations show the close relationship between platforms and service identity. For example, innovation in military aviation became easier when air branches became independent organizations, separate from armies and navies.119 Similarly, operators of transport or rotary-wing aircraft have often struggled to obtain the necessary recognition and resources from within their respective military services. These dynamics also occur between surface and underwater services, as well as between combat and intelligence or between surveillance and reconnaissance units. It remains to be seen where AI will fit into the culture of military organizations, but algorithms, data and processors are unlikely to be in the ensigns of any military service, at least in the foreseeable future: the tense discussions about counting drone operations as flying hours or assigning medals to drone pilots bear testament to that.120 This is a small, but powerful lesson: in the foreseeable future, some military services could lack interest in, or display insufficient attention to, these domains.121 This is why an AI champion within NATO may be particularly important.122 There is an additional consideration: interoperability would be difficult without coordination among Allies. At the NATO level, a centralized body could play a particularly useful and effective role in this respect. Future discussions could determine whether such a Centre should be independent, like a Centre of Excellence, or sit within the NATO Enterprise structure (natural options being under Allied Command Transformation or in the NATO Communications and Information Agency). There are, however, strong reasons to assign such a Centre a number of specific goals. Lead. The Centre should be at the forefront of NATO’s and the Allies’ AI efforts, including discussions about ethics and ownership of pilot projects (see below) as well as development of targeted solutions to existing problems and challenges. Support. The A3IC should have an interest in helping Allies adopt AI through a set of procedures, roadmaps, best practices and, where possible, readily available solutions, either developed in-house or borrowed from others. From ethics to training, from recruitment to procurement, from cyber security to data management, the Centre could provide important support, especially for some Allies or some of their services that may lack in- house solutions or expertise. Cases in point would be Testing and Evaluation (T&E) and Verification and Validation (V&V): with the adoption of ML, these will have to be rethought, upgraded and updated in order to integrate the non-deterministic nature of algorithms into existing procedures and methods.123 Similarly, in the age of software, procurement needs a major upgrade. In contrast to traditional military platforms, it is faced with a paradigm shift: rather than deliver finished and well-functioning products, it must come up with adaptable solutions that, by their very nature, can never be considered as “done” once and for all.124 Realistically, not all Allies possess the necessary expertise to face such challenges. The Centre could play an important role in these areas. Since the AI ecosystem, as this document shows, is admittedly massive but also scalable, for many Allies it may be much more convenient to rely on common, Alliance-wide capabilities, since they would probably struggle to achieve the required depth and the breadth if left to their own devices. Connect. The Alliance will not be effective if individual Allies’ efforts remain disconnected; and it will be more efficient if Allies are able to build on each other’s progress and achievements. The Centre could play a major role in this respect. For instance, the software community around the world relies on platforms such as GitHub to accelerate software development.125 The Alliance has the opportunity to move in the same direction, enabling Allies to benefit from each other’s progress. Researchers in different fields need the same types of tools, from science workflows to AI-driven experimentation, from testing and evaluation to other domains. If NATO’s A3IC could provide a central repository of AI software, it would connect all the actors, accelerate their work and also address potential failures.126 Similar considerations apply to data. “Usually, the biggest challenges [are] related to getting sufficient high-quality training data”. In fact, “system performance is directly tied to data quantity, quality, and representativeness”.127 For the A3IC, a key goal would be to make training data available, as this would dramatically accelerate Allies’ progress. Recent progress in AI has been possible, because nowadays there are “many open source code libraries and developer tools that allow organizations to use and build upon the work of external communities. As a result, no team or organization has to start from scratch, and may parts that used to require highly specialized expertise have been largely automated”.128 The A3IC could play a similar role in making all tools and tests available. This is particularly important for T&E/V&V activities, as will be discussed later.