Podcast Atlas
source Episode summary Updated 2026-07-24 Tags: Podcast, Innovation, Goals, Ai, Reading

139. 为什么伟大不能被计划：有志者事竟成，吗？

Summary

This [[MihuanChishu|蜜獾吃书]] episode uses [[WhyGreatnessCannotBePlanned|《为什么伟大不能被计划》 / Why Greatness Cannot Be Planned]] to question goal-first thinking in complex domains. The episode contrasts simple tasks, where goals and metrics can work, with research, careers, creativity, social policy, and reading, where Objective Deception can make the most direct-looking path a trap. Its strongest synthesis is that people and institutions should protect Novelty Search and Stepping-Stone Exploration: open exploration, interesting intermediate discoveries, and wider possibility space can matter more than proving constant progress toward a predefined endpoint.

Key Claims

  • The episode begins from the question of why to read if reading does not directly raise income, then reframes reading as Novelty Search rather than as a guaranteed income tool.
  • The source presents Kenneth Stanley and Joel Lehman as the AI researchers behind [[WhyGreatnessCannotBePlanned|《为什么伟大不能被计划》]], and uses that framing to import algorithmic and evolutionary examples into life-design questions.
  • Goal thinking works better for bounded tasks such as tidying a bookshelf than for open-ended problems such as scientific breakthroughs, career direction, cancer treatment, artificial intelligence, or social governance.
  • Objective Deception appears when an imperfect target function creates a false gradient: steps that look closer to the target may be dead ends, while useful intermediate steps may initially look unrelated.
  • The image-breeding experiment shows why complex outcomes are hard to reverse-plan. Images that later resemble castles, birds, or vehicles can descend from intermediate forms that did not resemble the final object.
  • The robot maze example is the episode’s clearest algorithmic contrast: according to the source, Novelty Search found the target in 39 of 40 complex-maze runs, while objective-driven search succeeded only 3 times.
  • Biological evolution is used as a non-teleological model: eyes, wings, lungs, and human intelligence are framed as products of constraints and accumulated variation rather than as preselected endpoints.
  • Stepping-Stone Exploration becomes a personal method: career shifts, job applications, volunteer ties, editing work, podcasting, and reading all become ways to find the next reachable possibility rather than execute a master plan.
  • The Richard Wiseman newspaper experiment is used to argue that over-focusing on the assigned task can make a person miss obvious opportunities in peripheral information.
  • Campbell’s Law names the policy and organization risk: when quantified indicators become targets, people and systems can optimize the number while corrupting the underlying purpose.
  • The episode does not reject all goals or metrics. It preserves goals for small tasks, fairness-constrained tests, and low-impact measurement while warning against overconfidence in major complex systems.
  • Basic research and seemingly useless projects are defended as necessary possibility-space expansion, not as automatically valuable outputs.

Key Quotes

“错误的指南针可能比没有指南针更危险” - the episode’s formulation of bad objective functions in complex domains.

“不如保持开放” - the closing practical stance toward exploration, reading, and discovery.

Connections

Contradictions

  • No direct contradiction found. The source qualifies, rather than rejects, existing pages that value goals, metrics, or action: it treats measurement as useful in bounded contexts while warning that complex open-ended systems can be harmed when a proxy target becomes the whole map.