We Need an International Treaty to Ban Superintelligence
The author argues that superintelligence poses an existential threat and calls for an international ban on its development.Summary
As AI companies race to create superintelligent systems, the author warns that current government measures like pre-deployment reviews and export controls are insufficient. Unlike nuclear weapons, superintelligence would be an autonomous agent that cannot be controlled by humans, representing a catastrophic principal-agent problem. The author contends that an international prohibition is necessary because the technical challenges of controlling such systems are far from being solved. They point out that the industrial process behind advanced AI, including chip fabrication and data centers, offers a path for monitoring and enforcement. The author urges the United States to lead an international agreement, similar to nuclear nonproliferation, to prevent the development of superintelligence before the window to act closes.
(Source：Google News)