This repository contains two documents commissioned by Open Philanthropy, on using forecasting to predict AI progress.
- AI safety forecasting questions OpenPhil might find useful is a big list of around 700 AI forecasting questions. Its folder also contains a few adjacent documents, like an operationalization of a resolution council, a discussion of how to operationalize FLOPs, and a small program to count the number of questions of each type.
- Hurdles of using forecasting as a tool for making sense of AI progress outlines some problem of using forecasting to predict AI progress. For convenience, a pdf version is provided here. It is also available as a blogpost, with a comment section, here.
This list of forecasting questions was originally developed by David Mathers, Gavin Leech and Misha Yagudin, of Arb Research and then completed by Nuño Sempere, of Shapley Maximizers. Open Philanthropy provided funding. The hurdles document was primarily written by Nuño Sempere, with input and ideas from Misha Yagudin.