Free beta access. 50 free AI tokens each week. Optional top-ups available. See what is included
Opening Word document…

Market forces can’t solve the aff---corporate r&d can’t meet NATOs needs on their own

Andrea Gilli et al 20, Senior Researcher at the NATO Defense College where he works on issues related to technological change and military innovation. He has been visiting and postdoctoral fellow at Johns Hopkins University and Columbia University as well as Stanford University (where he remains an Affiliate) and Harvard University. Mauro Gilli is a Senior Researcher in Military Technology and International Security at the Center for Security Studies (CSS) at the Swiss Federal Institute of Technology, ETHZurich. Before joining CSS, he was a post-doctoral fellow at the Dickey Center, Dartmouth College. Ann-Sophie Leonard is a former Mercator Fellow on International Affairs, focusing on the intersection of international security and technology. She was a Visiting Fellow at the Research Division of the NATO Defense College in Rome from September to December 2019. Zoe Stanley-Lockman is an Associate Research Fellow in the Military Transformations Programme at the Institute of Defence and Strategic Studies at the S. Rajaratnam School of International Studies in Singapore. “NATO-Mation”: Strategies for Leading in the Age of Artificial Intelligence. NDC Research Paper No.15 – December 2020 //pipk

Commercial technology and public goods. A common theme in discussions about AI is the role of private industry. The story goes that the commercial sector is leading in innovation on AI, and that the defence establishment, in contrast to the Cold War, thus needs to adapt to the needs and capabilities of the private sector, rather than the other way around.240 While defence establishments would certainly do well to shed some bureaucratic deadweight that constrains them from following the “lean”241 or “agile”242 ways of the commercial sector, it is problematic to import private-sector assumptions and practices as this line of thinking suggests.243 Namely, market forces do not always and necessarily lead to the social optimum: markets sometimes lead to the under-provision of public goods such as national defence, because of the inherent difficulty of charging all potential beneficiaries. Similarly, markets may favour the overproduction of negative externalities like pollution, because producers bear only a tiny fraction of the total cost. We observe these dynamics of market failure when we look at AI-related R&D by private companies. First, corporate R&D generally has a narrow, short-term focus, to the detriment of broader, longerterm issues which are necessary for ensuring that AI yields positive returns in the distant future.244 Second, the computer intensiveness of ML is technically, economically and environmentally unsustainable.245 In particular, nobody has mastered how to properly size ML models and modern deep-learning algorithms are based on over-parametrization. This means their performance is inherently inefficient, in that they need an enormous amount of data and parameters. The cost comes in terms of computing resources. Realistically, existing trends are not sustainable.246 NATO and its Allies have operational and strategic incentives to promote leaner and greener AI. Operationally, there is a limit to the utility of computationally intensive AI systems “at the edge” – i.e., usage by small units in operations where connectivity is not assured.247 Strategically, correcting cases of market failure enhances technological prowess insofar as it means promoting broader, longerterm research on AI and the development of less computationally intensive solutions not only for ML, but also for other technological priorities such as quantum248 and aspects of biotechnology. Moreover, weaning ML models off massive training data is one measure of AI becoming more robust,249 and thus potentially less susceptible to data-driven biases.250 Procurement. R&D will, sooner or later, lead to procurement programmes. Transitioning AI into prevailing procurement paradigms, however, will not be immediate or easy. Defence procurement is particularly complex, given the range of strict requirements and specifications with which contractors have to comply.251 This is why traditional weapons manufacturers generally have an advantage when it comes to approaching defence buyers.252 AI, however, will entail a paradigm shift because the capabilities will continuously evolve. For AI-enabled autonomous systems in particular, as mentioned above, decision-making is “non-deterministic” and depends on the “dynamic environment” in which the system operates.253 This means that “traditional development and procurement approaches, based on full-path regression, are unfit”.254 This means that R&D also continues in the product deployment phase, as data and algorithms keep providing information and feedback that have to be integrated in order to achieve initial operational capabilities. Current “waterfall” procurement paradigms are set up so that engineers test prototypes according to defined specifications, and then subsequently move to production.255 With machine learning, the specifications will keep evolving as algorithms are fed new data.256 By extension, this means that testing cannot be treated as a singular phase prior to production and development. For NATO Allies, a more iterative procurement paradigm that does not depend on the sequential “waterfall” entails unique challenges. The integration of enterprise AI requires specific technical, legal, and organizational capabilities. This is linked to human-capital challenges in requirements, procurement and operations communities, as organizations leveraging AI “need their own people who know how to structure the problem, handle the data, and stay aware of evolving opportunities”.257 The development of AI-centred major weapon systems similarly calls for reform, or at least a remarkable adaptation, of defence procurement.258 Significant political, organizational and human capital will have to be invested. This is an additional reason why a centre such as the proposed A3IC could support the transition, through best practices, lessons learned and similar initiatives. Supply chain. Ben Buchanan has conceptualized AI inputs as a triad: algorithms, data and processors.259 Processors are the material part of the triad and they pose a key supply chain vulnerability.260 NATO as an organization has neither the capabilities nor the authority to strengthen the AI supply chain. Additionally, the supply chain is heavily skewed towards some Allies, like the US and Great Britain in semiconductors or the Netherlands in semiconductor manufacturing equipment, and partners, like Israel, South Korea and Japan – plus Taiwan, which is a close ally of several NATO countries.261 A useful exercise for the Alliance could be to make an assessment of the supply chain related to semiconductors. This is important for three main reasons. First, several Allies are conducting their own assessments, but an Alliance-wide picture is so far lacking.262 Second, as several countries consider semiconductors a strategic industry and thus aim at achieving a degree of selfsufficiency and autonomy, there is a real risk that different national strategies may collide, thus leaving the Alliance as a whole worse off. Last, but not least, such an assessment could be helpful in determining potential joint or common solutions, including arms control and export controls, as discussed in the next chapter, as well as coordination in R&D investments.

Sign in to recut, update, or save this card.