Getting the G out of AGI

Assumptions about the inevitability of artificial general intelligence, or AGI, underpin the vast majority of today’s AI narratives. These claims about machines that will match and then quickly surpass and replace human intelligence, which once had the cast of speculative fiction, have now gained secure purchase in the minds of policymakers around the world who routinely invoke the necessity of preparations for AGI’s expected arrival. Many now speak as if they have no higher task. This talk challenges the intellectual viability and conceptual coherence of the very concept of general intelligence, and urges the scientific community to help delegitimise its use in both research and policy, much as the concepts of phlogiston and the aether were eventually pushed outside the bounds of scientific legitimacy. Only then can wider public understanding of AI be reestablished on sound footing, enabling leaders, journalists, and policymakers to shift from asking ‘When will AGI arrive?’ (a bad scientific question, and a harmful policy one) to a much more important question: ‘What cluster of intelligent capabilities (human and nonhuman) do people and the planet actually need in order to survive and flourish, that they do not already have?’