A proposed frontier AI training pause using chips locked to specific models

A researcher shared a 210-page analysis of whether a durable pause on frontier AI training is feasible. They say it proposes hardwiring specific models into chips so the chips can only run those models, rather than dismantling the AI compute supply chain. They argue existing models could still be deployed at scale, limiting the immediate economic hit, but worry that unknown model capabilities and new ways of using models could leave risks.