Ethereum Researcher Dankrad Feist Calls for a Pause on Training Cutting-Edge AI Models

Techub News reports that Ethereum Foundation researcher Dankrad Feist (@dankrad) posted on the X platform calling for a pause on training cutting-edge AI models. He stated that current AI models are already powerful enough and society should spend ten years fully absorbing the existing capabilities. Continuing to push the frontier would significantly increase risks, and humans need to control the pace of AI development within a time scale that humanity can handle. He proposed a specific implementation plan for the pause, including that the United States appropriately prohibit the training of new cutting-edge models, and that the US and China reach a treaty while encouraging more countries to join. On the execution level, restrictions could be implemented by limiting hardware or large data centers. He emphasized that the pause is not permanent; the goal is to allow societal development to catch up, enhance resilience, and consider the next steps after a few years.