An AI safety perspective grounded in data, mechanisms and policy trade-offs

A writer argues for moving AI safety debates from abstract claims toward data and concrete mechanisms, rather than ungrounded thought experiments. They point to past technological shifts and economic, policy and social feedbacks as useful context for assessing runaway harms. The writer says this approach does not dismiss catastrophic risks.