Long term cost-effectiveness of resilient foods for global catastrophes compared to artificial general intelligence safety

Journal article

Global agricultural catastrophes, which include nuclear winter and abrupt climate change, could have long-term consequences on humanity such as the collapse and nonrecovery of civilization. Using Monte Carlo (probabilistic) models, we analyze the long-term cost-effectiveness of resilient foods (alternative foods) - roughly those independent of sunlight such as mushrooms. One version of the model populated partly by a survey of global catastrophic risk researchers finds the confidence that resilient foods is more cost effective than artificial general intelligence safety is ∼84% and ∼98% for the 100 millionth dollar spent on resilient foods and at the margin now, respectively. Another version of the model based on one of the authors produced ∼93% and ∼99% confidence, respectively. Considering uncertainty represented within our models, our result is robust: reverting the conclusion required simultaneously changing the 3-5 most important parameters to the pessimistic ends. However, as predicting the long-run trajectory of human civilization is extremely difficult, and model and theory uncertainties are very large, this significantly reduces our overall confidence. Because the agricultural catastrophes could happen immediately and because existing expertise relevant to resilient foods could be co-opted by charitable giving, it is likely optimal to spend most of the money for resilient foods in the next few years. Both cause areas generally save expected current lives inexpensively and should attract greater investment.

Authors: Denkenberger, D; Sandberg, A; Tieman, RJ; Pearce, JM

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Elsevier
• Journal: International Journal of Disaster Risk Reduction
• Volume: 73
• Article number: 102798
• Publication date: 2022-02-23
• Acceptance date: 2022-01-14
• DOI: 10.1016/j.ijdrr.2022.102798
• ISSN: 2212-4209
• Language: English
• Keywords: FFR; nuclear winter; artificial intelligence; global catastrophic risk; cost-effectiveness analysis; existential risk; food security