Andy Burnham used his recent speech at the UN General Assembly to call for “a single set of global principles and standards” to govern AI. At the same meeting, Finland corralled 22 countries into signing a joint statement demanding AI be kept safely “under human control.”
It is perhaps not so surprising that left-wing leaders, and smaller countries left behind in the AI race, want to regulate technology. But the range of voices calling for regulation is much larger. Ezra Klein, the New York Times columnist and co-author of Abundance, recently came out in favour of a ban on recursively self-improving AI. A few days later, Francis Fukuyama announced that he had “changed his mind” on AI risk, writing that “regulation and a negotiated slowdown in the development of the technology is increasingly necessary.”
In the debate between AI ‘accelerationists’ and AI ‘doomers’, both Fukuyama and Klein appear to have sided with the doomers. But what’s interesting is that neither is a typical doomer: Klein is known for opposing burdensome regulations and accelerating technological development to support economic growth. Fukuyama has also been critical of regulatory constraints on the economy, and admits that he was sceptical of AI doomer scenarios until recently.
What changed? Recent events have rightly spurred calls for regulation at the frontier of AI. But for the UK, the calculation is different. Since we do not have firms at the frontier of AI, unilateral regulation makes little sense. We should instead prioritise model access for the AI Security Institute (AISI), and do everything we can to increase our AI leverage. This means that the optimal UK AI policy, from the perspective of global AI safety, may even look ‘accelerationist’.