About A Narrow Path
What is A Narrow Path?
A Narrow Path is a series of principles for policies that humanity should implement to survive AI, at least for the next 20 years. A Narrow Path aims to provide a coherent worldview and strategy to address extinction risk from AI: it is not a static online essay, but a project iterated over time.
Rather than specific laws, which can be highly contextual depending on jurisdiction, A Narrow Path details guiding principles to help policymakers develop solutions, and weigh other policy proposals.
This project is hosted on a website where the most up-to-date version is always available, with past versions archived and a changelog documenting major edits, changes and new contributors.
Why is "A Narrow Path" needed?
To tackle any problem in any domain, first the problem must be identified, then a strategy is drawn, then concrete implementation begins: this process is at the core of systems engineering, military strategy, and (at least in theory) public policy.
The three steps are: Problem identification → Strategy definition → Nitty-gritty implementation.
A Narrow Path is the “Strategy definition” part of this process. The problem is defined and discussed elsewhere: for the purposes of this project, it is “extinction risk from artificial intelligence”. If you disagree with the existence of the problem, A Narrow Path is not for you.
While the problem identification stage has already happened, we found no comprehensive Strategy on extinction risk from AI, even though there have been some individual policies proposed over time. So we made a strategy, and here it is.
A Narrow Path should answer the question: given the problem “extinction risk from AI”, what is the best course of action to address it, and the general principles that will ensure the problem is solved? A Narrow Path should provide the principles that allow people to generate concrete policies and laws that directly solve this problem, the “nitty-gritty implementation” part of the pipeline.
Authors
Contributors
Thank you to Anthony Aguirre, Connor Leahy, Max Tegmark, Eva Behrens, Leticia Garcia Martinez, Gabriel Alfour, Adam Shimi, Pedro Serodio, and many others for contributions and vital discussions. Thank you to everyone else that provided feedback on drafts. Thank you to Eleanor Gunapala for graphics and publishing.