Answers are grounded in this paper and cite their sources.
This record pre-registers nine interim predictions (M1–M9) and two conditional sub-predictions (M1.1, M4.1) on the governance of advanced AI. They are derived from the author’s scenario analysis Window, Oscillation and Lock-in: Scenarios and Long-Run Evolution of AGI Governance (2026–2040) (preprint V1.0, in Chinese). Each item states an event that can be adjudicated by 31 December 2027, 30 June 2028 or 31 December 2028. For each event the registration gives explicit criteria and data sources, a single probability fixed at registration, the conditions under which it is not adjudicable, and evaluation by the Brier score. The items cover: a US–China frontier-AI agreement with verification clauses, and whether it contains an anti-transfer clause; a pause mechanism in the bilateral AI incident channel; reliable measurement of AI time horizons of 40 hours or more; a major cross-border event that two governments formally attribute mainly to AI; laboratory statements that a named R&D stage is more than half autonomous; occupational employment declines in an OECD country; permission-before-completion for irreversible actions by AI agents; multilateral verification, either as a club-type agreement or as an independent body tied to chokepoint supply. Only public formal texts count, and vague or merely intentional wording counts as not holding. The probabilities are the author’s conditional subjective judgements (tier C), not outputs of the underlying framework. Eleven items cannot establish calibration, and items that share sources with the earlier registration P23 count once. The registration seals off ex post adjustment. It does not certify the framework, and its adjudication is not a full audit of reality. The English text is authoritative.
No takes yet. Share an insight, caveat, or question.
Qinfu Li (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: