5D Library › 五维灵商量化架构、先验条件化生成与检索增强认知体 · English

第七章 仰观俯察与邵雍观物

为什么在这里读这些文献

本书的三个问题,即一个人可以从几个互相独立的方面来观察,一套分类能分出多少类,分类是否真的能预测什么,在中国思想中都有先例。本章读两份文献:《周易·系辞上》论「仰观俯察」的一段,与当代学者 Peter K. Bol 对邵雍观物之法的研究。《系辞》引文已与所列底本逐字核对;Bol 的文字引自其文稿的英文原文,同样逐字核对。

《系辞上》:仰以观于天文,俯以察于地理

原典

《周易·系辞上》(中国哲学书电子化计划所录本)

易與天地準,故能彌綸天地之道。仰以觀於天文,俯以察於地理,是故知幽明之故。原始反終,故知死生之說。

(Anonymous, n.d.)

这段话提出了一种认识方法:向上观察天象的纹理,向下考察大地的条理,由此知道幽隐与显明的缘故;追溯事物的开始、反观它的终结,由此知道生死的道理。用本书的语言说,这是两个互相独立的观察方向。第一章的计算表明,独立的维度之所以有价值,正是因为它们不能互相替代:若「仰观」与「俯察」看到的是同一件事(维度相关很高),第二个方向就几乎不增加信息。程序第三项给出最简单的计数:若每个方向单独能分出 mm 类,两个独立方向合起来能分出 m2m^2 类。

邵雍的观物之法

原典

Peter K. Bol, “On Shao Yong’s Method for Observing Things”(英文文稿,逐字引用)

the work that we now have that is accepted as being from Shao Yong’s own hand

dividing any topic into fours aspects

the number of shi of yuan is 1x12x30x12=4320

a well-organized and methodologically consistent treatise

(Bol, n.d.)

据 Bol 的研究,《观物内篇》是今天公认出自邵雍本人之手的著作;邵雍把任何一个题目都分为四个方面来观察,而《皇极经世》中一元所含的「世」数为 1×12×30×12 = 4320。Bol 称《观物内篇》是「一部组织严密、方法一贯的论著」。(第一句引文中的「fours aspects」依原文照录。)

推导

# Chapter 7. p14: The numbers of Shao Yong's method, recomputed.
# (1) Bol's reading of the Huangji jingshi: one yuan holds 12 hui, a
# hui 30 yun, a yun 12 shi, so a yuan holds 1 x 12 x 30 x 12 = 4320
# shi. If a shi is taken as 30 years (the usual sense of the word), a
# yuan spans 4320 x 30 years.
# (2) Bol: Shao Yong divides any topic into four aspects. Dividing k
# times gives 4^k classes, 2k bits. Set against the profile of p05:
# five dimensions of three levels, 3^5 = 243 states, log2 243 bits.
# (3) The Xici pairs looking up at the patterns of heaven with looking
# down at the order of the earth: two independent views. If each view
# alone separates m kinds of things, the two together separate m^2.

import math

if __name__ == "__main__":
    shi = 1 * 12 * 30 * 12
    print("   (1) shi in a yuan: 1 x 12 x 30 x 12 = {:,d}".format(shi))
    print("       years at 30 a shi: {:,d}".format(shi * 30))
    print("   (2) k   classes 4^k   bits")
    for k in range(1, 6):
        print("       {} {:12,d} {:6d}".format(k, 4 ** k, 2 * k))
    print("       profile of p05: 3^5 = {} states, {:.4f} bits".format(
        3 ** 5, math.log2(3 ** 5)))
    print("   (3) one view m kinds -> two views m^2: " + ", ".join(
        "{} -> {}".format(m, m * m) for m in (4, 8, 16)))
    print("\nReading: four successive fourfold divisions give 256")
    print("classes, eight bits - almost exactly the 7.92 bits of a")
    print("five-by-three profile. Classification schemes of this size")
    print("are easy to build; what neither Shao Yong's tables nor a")
    print("JSON profile can show by themselves is that the classes")
    print("predict anything (p05, p06).")
#    (1) shi in a yuan: 1 x 12 x 30 x 12 = 4,320
#        years at 30 a shi: 129,600
#    (2) k   classes 4^k   bits
#        1            4      2
#        2           16      4
#        3           64      6
#        4          256      8
#        5        1,024     10
#        profile of p05: 3^5 = 243 states, 7.9248 bits
#    (3) one view m kinds -> two views m^2: 4 -> 16, 8 -> 64, 16 -> 256
#
# Reading: four successive fourfold divisions give 256
# classes, eight bits - almost exactly the 7.92 bits of a
# five-by-three profile. Classification schemes of this size
# are easy to build; what neither Shao Yong's tables nor a
# JSON profile can show by themselves is that the classes
# predict anything (p05, p06).

程序先复算邵雍的数:一元 12 会、一会 30 运、一运 12 世,共 4,320 世;若按「世」的通常含义取一世 30 年,一元为 129,600 年(这是按参数计算的结果,不是对《皇极经世》原文的引述)。再算四分法:连续分四次得 256 类,即 8 比特;而第三章的五维三档档案有 243 种状态,7.9248 比特,两者几乎相同。

这说明,构造一个分类体系并不难:四次四分与五维三档给出的类数相差无几。难的是证明这些类能预测什么。邵雍的表格与一份 JSON 档案在这一点上面对的是同一个问题;第三章的计算表明,答案取决于回答是否真的依赖分类,而不是分类本身有多精细。

假说 · 尚未验证

作者的诠释(尚未验证): 邵雍的「观物」与《系辞》的「仰观俯察」都是多维、互相独立的观察框架,五维灵商是这一传统的现代量化形式。什么会证明这一诠释不当:若第一章所述的维度相关在真实数据中很高(五维几乎是一维),那么五维框架在测量上并不比单一维度提供更多的「观察方向」,这一类比只在形式上成立。

小结

《系辞》的「仰观俯察」提出了两个独立的观察方向;邵雍把任何题目分为四个方面,并以 1×12×30×12 = 4320 推算一元之世。四次四分得 256 类、8 比特,与五维三档档案的 7.9248 比特几乎相同。分类容易,证明分类能预测什么才是难处。

参考文献

Anonymous. n.d. Zhouyi, Xici Shang (Appended Statements, Part One). In Chinese; Chinese Text Project (Ctext.org).
Bol, Peter K. n.d. On Shao Yong’s Method for Observing Things. Author’s Manuscript, Harvard University, DASH Repository.

Fan, Y. (2026). Spiritual Quotient Quantification, Prior-Conditioned Generation and Retrieval-Augmented Cognitive Agents: Multidimensional State Vectors, Personalized Prompt Conditioning and Cross-Corpus Synthesis [五维灵商量化架构、先验条件化生成与检索增强认知体] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090203 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).