FRI Forecasts 0.48% AI Catastrophe Risk by 2030, Impacting Tenth of Humanity

The Forecasting Research Institute (FRI) has estimated a 0.48% chance of an AI-related catastrophe by 2030. This projection defines a catastrophic incident as one leading to the demise of at least one-tenth of humanity. FRI's methodology involved aggregating forecasts from 25 AI experts, each contributing an independent probability assessment.

These experts engaged in structured forecasting processes during the summer of 2023. They were tasked with predicting various scenarios regarding AI's impact. The institute specifically focused on extreme negative outcomes.

The FRI's approach leverages superforecasting techniques pioneered by Philip Tetlock. This method aims to improve forecasting accuracy through structured elicitation and aggregation. The institute is based at the University of Pennsylvania's Wharton School. It combines insights from cognitive psychology and data science.

Funding for FRI's work includes contributions from Open Philanthropy. This support facilitates its ongoing research into long-range societal risks. The institute's findings provide a quantitative data point for risk assessment. Its mission involves applying rigorous forecasting methods to critical future events. This includes developing and testing new forecasting tools.

How might such precise, low-probability forecasts influence policy decisions?