We Won't Know the Answers to AI's Most Important Questions Until It's Too Late

TIME op-ed (3 Oct 2026) by Geoffrey Irving, described as former chief scientist at the UK AI Security Institute and current chief scientist of Resolution. He writes: "I believe there's about a 50% chance we all die because of the development of smarter-than-human AI systems". He says actions over "the next two to 10 years will determine the outcome". He names hacking, persuasion, deception and coordination as the capabilities through which superintelligent AI could kill humanity. He argues that the key debates will stay unresolved until it is too late to change course. These are whether skills transfer across tasks, whether AI will exceed human level, whether small-scale misbehaviour scales with capability, and whether current safety methods work on superintelligent systems. His conclusion is that society must act under uncertainty. He advocates an immediate pause on frontier development, with international cooperation modelled on nuclear non-proliferation. He notes that US "support for a pause on AI development has increased tremendously over the past few weeks". Read via WebFetch, so quotes come through a summarizer. Verify the exact wording against the page.

Short quotes only (see Summary). Full text at the URL.