Great power conflict

Abstract

This work examines how transformative AI, capable of causing a 10X acceleration in economic growth, can be forecast. Simpler forecasting methods are based on extrapolating curves fitted to historical data or growth rates, or obtaining expert opinions through surveys. A more AI-centric methodology presented here is through what are called biological anchors, which start from the premise that an AI system may need as much computation to operate as the human brain itself. This involves estimating how much computation would be needed to train a human-level AI model and using trends in compute hardware, software, and resources to forecast when this level of computation might be reached. – AI-generated abstract.