uses strong AGI from this Metaculus question: https://www.metaculus.com/questions/5121/date-of-artificial-general-intelligence/

pause doesn't have to be continuous but cumulative pause time

Resolution criteria

This is a conditional market. It will resolve to N/A if a global pause on frontier AI training does not occur before "strong AGI" is achieved.

If a global pause on frontier training is enacted prior to the arrival of strong AGI, this market will resolve to the option representing the cumulative duration of the pause(s) before strong AGI is officially achieved.

• Strong AGI is defined as satisfying the resolution criteria of Metaculus Question 5121: Date of Artificial General Intelligence (which requires a single unified software system to pass a 2-hour adversarial Turing test, assemble a scale model car in robotics, score high on MMLU/APPS benchmarks, etc.).

• A "Global Pause" is defined as an internationally recognized, legally binding, or highly coordinated voluntary cessation of all frontier training runs (e.g., via multilateral treaty, coordinated national regulations/executive actions, or joint agreement among all leading frontier AI labs including OpenAI, Google DeepMind, Anthropic, Meta, and major Chinese developers).

• Frontier Training refers to training runs of state-of-the-art AI models exceeding advanced compute or capability thresholds (such as $>10^{26}$ FLOPs, or those subject to regulatory reporting requirements at the time).

• Cumulative Duration is the sum of all individual pause periods; the pause does not need to be continuous to count toward the total.

• Resolution Source: Resolution will be determined based on credible journalistic reporting, official treaty or government documents, and the ultimate resolution timeline of Metaculus Question 5121. In the case of minor boundary disputes or ambiguous pause definitions, the market creator will resolve the market based on consensus from reputable AI governance and research organizations (e.g., Epoch AI, Future of Life Institute).

Options:

• Less than 6 months

• 6 to 24 months

• 2 to 5 years

• 5+ years

Background

The concept of a "frontier training pause" became a major topic of debate in AI governance following the March 2023 Future of Life Institute open letter, which proposed a temporary cessation of training models more advanced than GPT-4. Since then, international discussions and agreements (such as those stemming from the global AI Safety Summits) have continuously weighed "circuit breakers" and pausing mechanisms as potential safeguards against the rapid emergence of uncontrollable or catastrophic capabilities in next-generation systems.

This market evaluates how long such pauses might collectively last before a system meeting Metaculus' rigorous definition of "strong AGI" is finally developed. If AGI is achieved without any global pause ever being implemented, the precondition is not met and the market resolves to N/A.