It argues that much high-value work in science, business, public policy, medicine, and AI research falls into the latter category. The essay names proxy selection, theory-building, resource awareness, and error correction under scarce feedback as relevant capabilities. It presents a conceptual transfer problem, not a controlled result showing a particular model’s performance gap. Source: https://www.lesswrong.com/posts/2mStxuNDMdaWdyHu6/the-expensive-loop-problem
The Expensive-Loop Problem (22 minute read)
In one minute
LessWrong published *The Expensive-Loop Problem*, an essay distinguishing “cheap-loop” tasks, where experiments are quick, deterministic, and inexpensive, from “expensive-loop” tasks constrained by real-world cost or delay.