The asymptotic limit theorems of information theory, and related formalism from punctuated equilibrium evolutionary theory, allow exploration of the dynamics of real-time, multi-component, hierarchical, cognitive enterprises – institutional, machine, or man/machine hybrid – under constraints of time, material resources, internal communication, situational awareness, and adversarial intent. Surprisingly direct argument shows such enterprises can almost always be confounded by an opponent with comparatively limited resources, provided it has sufficient situational understanding, adaptive capacity, and resolve. Consequently, there can be no free lunch for AI combat systems, driverless cars on intelligent roads, or other such entities, tasked with the control of real-time critical processes. Intelligent adversaries, resource limitations, “design flaws”, inattentional blindness, command stupidity, coevolutionary dynamics, and the like confront all real-time, multi-component cognitive processes. AI systems will not be spared simply by being labelled “high tech” in marketing campaigns.

中文翻译：

---

竞争、摩擦和战争迷雾下的认知不稳定性挑战人工智能革命

信息论的渐近极限定理，以及来自间断均衡进化理论的相关形式主义，允许探索实时、多组件、分层、认知企业——制度、机器或人/机混合——在时间限制下的动态、物质资源、内部沟通、态势感知和对抗意图。令人惊讶的直接论据表明，只要有足够的态势理解、适应能力和决心，这些企业几乎总是会被资源相对有限的对手所迷惑。因此，对于人工智能作战系统、智能道路上的无人驾驶汽车或其他此类负责实时关键流程控制的实体来说，没有免费的午餐。聪明的对手，资源限制，“设计缺陷”、疏忽失明、命令愚蠢、协同进化动力学等都面临着所有实时的、多组件的认知过程。人工智能系统不会因为在营销活动中被贴上“高科技”标签而幸免于难。