Finding the right questions is crucial in order to build a repository of predictions which are not just accurate, but also useful to organisations making progress on AI safety and policy.
When making your suggestions...
- Feel free to be brief, hand-wavy, etc.
- We've found that in early stages the gist underlying a question is more important than the exact formulation, so make sure to mention why you care about your suggested question
- In case questions are promising, the Metaculus AI team has the bandwidth to help with operationalising, gathering background data and crafting good resolution conditions.
- In case something is hard to capture using only a single question, we can get at it using a "Bayesian question cluster" of multiple related questions. We're not really constrained by number of questions we can add at the moment.