龙序体系综述:从存续公理到语义重量

Abstract

摘要(中文) 本综述系统梳理以「存续先于存在」为元公理的龙序体系,涵盖其演进脉络、核心判据、与哲学史的对话,以及在人工智能时代的落地方案。 体系部分给出存续约束四大公理(边界锚定、负熵摄入、损耗管控、压力适应)的新创与承继逐条审计,指出其原创性在于重组、跨尺度统一与可操作化,而非每条的物理内容本身;并明确自创生仅为互构生的单系统特例,不构成体系根基。 哲学对话部分处理与巴门尼德、萨特、海德格尔、休谟、查尔默斯、维特根斯坦、奥斯汀、福柯、尼采、祖博夫及泰勒的关系,其中对休谟问题采取「不消解而划定适用范围」的立场,对查尔默斯困难问题采取「重框而非解决」并诚实标注边界。 核心判据部分提出语义重量的双轴结构(重量 W 与保真度 F),并区分两种不可逆性:热力学第二定律的过程不可逆(无主、破坏性、无方向)与语义重量的归属不可逆(有主、建设性、有方向);以擦除操作与个体经验的对偶案例证明二者互相独立,并以黑洞作为「过程不可逆而归属为零」的第三样本,由此提出跨尺度自相似:AI 之于人,即黑洞之于星系。 落地部分给出 A/B/C 三级归属判据、分润模型的前提(归属与计量的区分)、伪造举证的攻防,并明确该分级是成本机制而非检测机制。 文末列出体系自身的失效清单与开放项,包括涌现缺口、qualia 未解决、保真度尚未操作化等。方法论上,本体系自定位为自然化的形而上学:物理定律提供约束而非还原,归属维度属形而上学承诺而非物理推论。 Abstract (English) This survey systematically presents the Longxu framework, built on the meta-axiom "subsistence precedes existence," covering its development, core criteria, dialogue with the history of philosophy, and proposed applications in the age of artificial intelligence. The systematic part audits the four operating axioms of subsistence constraint (boundary anchoring, negentropy intake, loss management, pressure adaptation) item by item for what is inherited and what is new. It argues that their originality lies in recombination, cross-scale unification, and operationalization—not in the physical content of any single axiom. It also establishes that autopoiesis is only the single-system special case of mutual constitutive becoming, and therefore not the foundation of the framework. The philosophical part addresses the framework's relations to Parmenides, Sartre, Heidegger, Hume, Chalmers, Wittgenstein, Austin, Foucault, Nietzsche, Zuboff, and Taylor. On Hume's problem it takes the position of delimiting the scope of the is/ought gap rather than dissolving it; on Chalmers' hard problem it offers a reframing rather than a solution, and marks that boundary explicitly. The core criteria part proposes a two-axis structure for semantic weight (weight W and fidelity F), and distinguishes two kinds of irreversibility: the process irreversibility of the second law of thermodynamics (ownerless, destructive, directionless) and the attributional irreversibility of semantic weight (owned, constructive, directed). A pair of opposing cases—bit erasure and an individual's accumulated experience—shows the two to be independent. A black hole serves as a third sample, in which process irreversibility holds while attribution is zero, yielding a cross-scale isomorphism: AI is to humans as a black hole is to a galaxy. The application part sets out a three-level attribution criterion (A/B/C), the presupposition of any revenue-sharing model (the distinction between attribution and measurement), and the attack-and-defense analysis of forged evidence, stating explicitly that the classification is a cost mechanism rather than a detection mechanism. The survey closes with the framework's own failure list and open problems, including the emergence gap, the unresolved status of qualia, and the not-yet-operationalized notion of fidelity. Methodologically, the framework positions itself as a naturalized metaphysics: physical laws supply constraints rather than reductions, and the attributional dimension is a metaphysical commitment rather than a physical implication.