In its second iteration, the Singapore Defense Technology (SDTS) this week focuses on a key theme: the proliferation of technology and the impact of that proliferation on security, defense, and society. At the outset, controlling the proliferation of emerging technologies such as artificial intelligence (AI) – a key strategic technology area the Summit will explore – seems daunting. The cat may already be out of the bag, as the saying goes.
Many presume that the commercial nature of AI means it will automatically proliferate. On the one hand, there is the reason to believe that AI will be accessible to many – including to nefarious or adversarial actors. But conversely, there is also significant evidence that “AI democratization” arguments are overblown. While AI has the potential to expand the attack surface, it should not be seen as an equalizing force.
Be it to leverage a marketing advantage or laud impressive advancements in machine learning, AI seems to be everywhere. Organizations explicitly seeking responsible openness see sharing and collaboration as important values in the AI community. AI hype also characterizes the field as democratizing, particularly with pre-built AI “starter packs” of algorithms and user-friendly interfaces that make it possible for anyone − for better or for worse − to enter into the field. With such links, the narrative of AI democratization also connotes that access to AI is unfettered.
AI certainly does challenge traditional counter-proliferation techniques. Computer scientists have likened preliminary attempts to place export controls, a key mechanism to prevent the undesirable proliferation of critical technologies, on AI to “controlling the export of math” itself.
Because advances in AI come from modifications to a set of algorithms widely available in the public domain, there is a limit to how effective export controls can be. Unlike other sophisticated technologies, such as research on nuclear materials, there are no secrets that export control regimes can try to prevent from spreading.
Furthermore, in a strategic context, one second-order effect of emerging technologies is that the lower the cost, the higher the willingness to deploy it is. In particular, financial intensity, one of the key components to determining the diffusion potential of an innovation, typically translates to more experimentation. With technologies dependent on open-source information, such as AI, widespread interest and experimentation can, therefore, be expected.