ALIGNING REGULATION AND AI

In this paper for Google, Brian Williamson considers the potential of AI and the need to align regulation and AI.

AI is a general-purpose technology that like steam, electricity and computing, is expected to boost productivity growth, the only sustainable source of per capita income growth. AI will also help offset headwinds from an aging population, support innovation in areas including the life sciences and decarbonisation and improve safety in activities ranging from healthcare to transportation. 

AI is subject to existing laws and regulations. For example, if AI is utilised to identify a potential medicine, it will need to pass the same requirements in terms of efficacy and safety as existing drugs. It is therefore a false dichotomy to frame AI regulation as absent if not imposed via new law. The challenge is to adapt existing law and regulation over time, to align regulation and AI. 

AI will be utilised offensively and defensively. For example, AI offers the prospect of stronger cybersecurity defence, and potentially a reduction of harm overall. We should not therefoe limit the capability and availability of AI in ways that could constrain its defensive use.

A law of AI per se is unecessary, and likely to prove counterproductive. We did not introduce a law of steam at the outset of the industrial revolution, and innovation led to increases in the efficiency of steam engines and complementary developments such as steam locomotives over time. In response we introduced regulation – not of steam per se, but in relation to emerging issues such as rail and factory safety. This is the approach we should follow with AI, adapting rather than preempting.