Executive summary: The decision to deploy AI that could increase both economic growth and existential risk depends critically on the potential growth benefits, the existential threat, and the curvature of the utility function, with extensions showing that economic growth outside of AI deployment and the ability to pause AI to reduce risk can significantly impact optimal deployment decisions.
Key points:
The decision to deploy AI involves a trade-off between potential unprecedented economic growth and increased existential risk.
The optimal AI deployment time is mainly determined by the growth rate under AI, the existential risk posed by AI, and the curvature of the utility function (γ).
Higher potential growth from AI increases optimal deployment time, while higher existential risk decreases it. The impact of the utility function curvature depends on whether utility is bounded.
Allowing for economic growth outside of AI deployment generally reduces the optimal AI deployment time compared to the base model.
Pausing AI development to reduce existential risk can increase long-term welfare under certain conditions, such as low utility function curvature, low discount rates, or large risk reductions from pausing.
Extensions to the model, such as optimal pausing with uncertainty about AI risk and differential technological progress during a pause, could provide further insights for AI deployment decisions.
This comment was auto-generated by the EA Forum Team. Feel free to point out issues with this summary by replying to the comment, andcontact us if you have feedback.
Executive summary: The decision to deploy AI that could increase both economic growth and existential risk depends critically on the potential growth benefits, the existential threat, and the curvature of the utility function, with extensions showing that economic growth outside of AI deployment and the ability to pause AI to reduce risk can significantly impact optimal deployment decisions.
Key points:
The decision to deploy AI involves a trade-off between potential unprecedented economic growth and increased existential risk.
The optimal AI deployment time is mainly determined by the growth rate under AI, the existential risk posed by AI, and the curvature of the utility function (γ).
Higher potential growth from AI increases optimal deployment time, while higher existential risk decreases it. The impact of the utility function curvature depends on whether utility is bounded.
Allowing for economic growth outside of AI deployment generally reduces the optimal AI deployment time compared to the base model.
Pausing AI development to reduce existential risk can increase long-term welfare under certain conditions, such as low utility function curvature, low discount rates, or large risk reductions from pausing.
Extensions to the model, such as optimal pausing with uncertainty about AI risk and differential technological progress during a pause, could provide further insights for AI deployment decisions.
This comment was auto-generated by the EA Forum Team. Feel free to point out issues with this summary by replying to the comment, and contact us if you have feedback.