[PDF] Forecasting Existential Risks - Squarespace

Forecasting Existential Risks
Evidence from a Long-Run
Forecasting Tournament
Authors: Ezra Karger, Josh Rosenberg, Zachary Jacobs, Molly
Hickman, Rose Hadshar, Kayla Gamin, Taylor Smith, Bridget
Williams, Tegan McCaslin, Stephen Thomas, Philip E. Tetlock
First released on: July 10, 2023
Current version: August 8, 2023 |
FRI Working Paper #1

-- 1 of 754 --

1
Forecasting Existential Risks:
Evidence from a Long-Run Forecasting Tournament1
Authors: Ezra Karger,†,* Josh Rosenberg,* Zachary Jacobs,* Molly Hickman,* Rose Hadshar,*
Kayla Gamin,* Taylor Smith,* Bridget Williams,* Tegan McCaslin,* Stephen Thomas,*
Philip E. Tetlock‡,*
Abstract:
The Existential Risk Persuasion Tournament (XPT) aimed to produce high-quality
forecasts of the risks facing humanity over the next century by incentivizing
thoughtful forecasts, explanations, persuasion, and updating from 169 forecasters
over a multi-stage tournament. In this first iteration of the XPT, we discover points
where historically accurate forecasters on short-run questions (superforecasters) and
domain experts agree and disagree in their probability estimates of short-, medium-,
and long-run threats to humanity from artificial intelligence, nuclear war, biological
pathogens, and other causes. We document large-scale disagreement and minimal
convergence of beliefs over the course of the XPT, with the largest disagreement
about risks from artificial intelligence. The most pressing practical question for future
work is: why were superforecasters so unmoved by experts’ much higher estimates of
AI extinction risk, and why were experts so unmoved by the superforecasters’ lower
estimates? The most puzzling scientific question is: why did rational forecasters,
incentivized by the XPT to persuade each other, not converge after months of debate
and the exchange of millions of words and thousands of forecasts?
1 This research would not have been possible without the generous support of the Musk Foundation and the
Long-Term Future Fund. We thank Walter Frick, Michael Page, Terry Murray, David Budescu, Barb Mellers, Elie
Hassenfeld, Philipp Schoenegger, and Pavel Atanasov for their thoughtful feedback and comments. We are also
grateful for the many people who reviewed earlier drafts of this report. We greatly appreciate the assistance of
Amory Bennett, Kaitlyn Coffee, Adam Kuzee, Avital Morris, Fiona Pollack, Coralie Consigny, Arunim Agarwal,
and Rumtin Sepasspour throughout the project. Lastly, we extend our gratitude to our research participants for
their invaluable contributions.
† Federal Reserve Bank of Chicago
* Forecasting Research Institute
‡ Wharton School of the University of Pennsylvania

-- 2 of 754 --

2
Table of contents
1. Executive summary
Key takeaways
Abbreviated results
2. Background & motivation
3. How the XPT works
4. Results
5. Individual risk areas
Nuclear weapons
Artificial intelligence
Climate change
Biorisks
6. Next steps
7. Conclusion
Bibliography
Appendices
Appendix 1: Practical problem

... (truncated, 98 KB total)