Olle Häggström, Chalmers, Göteborg
AI risk, AI safety, and the challenge from AI successionism
Abstract
Ever since the field began in the 1950s, the ultimate vision of AI research been to create machines that match or even exceed human performance across the full spectrum of relevant cognitive capabilities. Already Alan Turing warned that if this succeeds, humanity might lose control to the machines. Over the years, such warnings have been mostly ignored due to expectations of long timelines until the critical breakthrough. However, due to recent very rapid advances in AI development causing many experts in the field to talk about timelines measured in single-digit years rather than decades or centuries, the issue has become more urgent, and the situation looks quite concerning. On the other hand, this conclusion has been challenged from various directions, including from advocates of so-called AI successionism, holding that the human species being replaced by advanced AI would actually be a good thing.

Senast uppdaterad: 2026-01-08
Sidansvarig: Statistiska institutionen