DeepSeek's AI: Is It Powered by Ancient Chinese Wisdom?
A hypothetical exploration of philosophical Influence
A few months ago, I read "Under Heaven - Tianxia" by Zhao Tingyang (Astrolabio Ubaldini Editore Publisher), translated into Italian by Alessandra Lavagnino, and yesterday it came back to my mind as I began reflecting on how this ancient system of thought might offer a surprisingly relevant interpretative key to understanding DeepSeek's success in the field of artificial intelligence. Tianxia (天下, "all under heaven"), developed during the Zhou dynasty (1046-256 BCE), represents a worldview that transcends mere political dimensions: it's a holistic system of thought that conceives cosmic, social, and political order as manifestations of a single universal harmony, in stark contrast to the competitive fragmentation of Western thought.
DeepSeek
DeepSeek's approach to technological development mirrors this integrated vision. The training of their DeepSeek-V3 model, with 671 billion parameters completed in just two months for $5.58 million, using only 2.78 million GPU hours (compared to Meta's 30.8 million required for Llama 3.1), embodies the Confucian principle of frugality (儉, jian). This concept, elaborated by Confucius in the Analects (論語, Lun Yu), is not simple material parsimony but a form of wisdom that sees moderation as the key to social harmony and operational efficiency.
The relational rationality of Tianxia manifests in DeepSeek's open-source strategy. While Western giants keep their models closed, DeepSeek has chosen to publish codes and models, offering APIs at 1 RMB per million tokens - approximately 1/53 of the cost of Claude 3.5 Sonnet. This choice reflects the concept of 關係 (guanxi), the interdependent relationships that in Chinese thought constitute the very fabric of social reality.
DeepSeek's response to American sanctions on advanced chips perfectly embodies the wisdom of the Tao Te Ching (道德經). As Lao Tzu writes in chapter 78: "Nothing in the world is softer and weaker than water, yet in attacking the hard and strong, nothing can surpass it" (天下莫柔弱於水，而攻堅強者莫之能勝). Deprived of access to H100 GPUs, they developed innovative solutions with H800 GPUs, demonstrating the same adaptability and resilience that Lao Tzu attributes to water. Their MLA (Multi-head Latent Attention) architecture, which reduces memory consumption to 5-13%, is a perfect example of how flexibility can overcome brute force.
Confucian principles
The DeepSeek team deeply embodies the Confucian principle of 修身 (xiushen, self-cultivation) elaborated in the "Four Books" (四書, Si Shu), particularly in the "Great Learning" (大學, Da Xue). 90% of the team consists of recent graduates from Tsinghua, Peking, and Fudan Universities, selected according to a philosophy reminiscent of the concept of 君子 (junzi, noble person) - not so much for technical skills as for what Confucius called 志 (zhi), the aspiration to moral and intellectual excellence.
This organizational structure also reflects the concept of 禮 (li, appropriate rituals) as described in the "Book of Rites" (禮記, Li Ji). It's not a rigid hierarchy but a natural order where, as Xunzi wrote in the 3rd century BCE, "harmony is not uniformity" (和而不同). True harmony doesn't require all elements to be equal or think the same way. On the contrary, just as music harmony arises from the combination of different notes, so in social and political organization, optimal harmony emerges from the coexistence and productive interaction of different elements, each maintaining its uniqueness while contributing to the whole. Young researchers have unlimited access to computational resources and autonomy in projects, creating what in classical Chinese thought is called 中和 (zhonghe), the dynamic balance between order and creativity.
The transmission of the "Mandate of Heaven" (天命, Tianming), a key concept developed by the Zhou dynasty, finds a striking parallel in DeepSeek's legitimation strategy. As documented in the "Classic of Documents" (書經, Shujing), the Zhou justified their power by inserting themselves into a narrative of historical continuity. Similarly, the choice of names like DeepSeek-Math deliberately evokes the heritage of the "Nine Chapters on Mathematical Art" (九章算術, Jiuzhang Suanshu), the fundamental text of ancient Chinese mathematics, creating what we might call a "tianxiacentric vortex" of cultural and technological legitimation.
Economic approach
DeepSeek's economic approach reflects the concept of 經世 (jingshi, world management) elaborated by Neo-Confucian thinkers of the Song dynasty. Their revolutionary pricing model - 1 RMB per million tokens - embodies the principle of 義 (yi, justice/appropriateness) in its economic aspect, as theorized by thinkers like Zhang Zai (1020-1077) who saw economics as an aspect of cosmic order.
CEO Liang Wenfeng, indirectly inspired by the concept of "大同" (datong, great unity) from the Book of Rites, has stated that the true competitive gap is not temporal but between "originality and imitation." According to Wenfeng, China cannot limit itself to following but must aim for innovation to achieve lasting leadership and contribute to a harmonious world order in which cultural and technical excellence determine the leadership role.
DeepSeek's technical innovations, such as the GRPO (Group Relative Policy Optimization) algorithm, reflect the Taoist concept of 無為而無不為 (wu wei er wu bu wei)—"non-action yet nothing is left undone"—achieving superior results with minimal resources through intelligent optimization rather than brute force.
Chinese strategic thought
This integrated approach, rooted in three thousand years of Chinese strategic thought, is potentially redefining the future of AI and the parameters of global technological competition. The combination of Confucian frugality, Taoist adaptability, and technological ambition offers an alternative model to Western hyper-individualistic capitalism. The ancient wisdom of the Tao Te Ching offers us an illuminating perspective through the verse "What is complete seems incomplete, yet its use is inexhaustible" (大成若缺，其用不弊).
This millennial paradox finds a surprising resonance today in the world of artificial intelligence through DeepSeek's work. What appears incomplete or imperfect in the world's eyes may harbour unlimited potential. It is often in the imperfections, voids, and limitations that true strength resides. While Western technology giants pursue ever larger and more complex AI models, DeepSeek has embraced a seemingly more contained approach that, just as the Taoist text suggests, transforms limitations into potential.
As François Jullien notes in *A Sage is Without Ideas* (1998) [Un sage est sans idée], "Chinese thought does not think about being, but about process" [La pensée chinoise ne pense pas l'être, mais le procès] – this reflection finds an echo in DeepSeek's approach, which transforms the wisdom of ancient Chinese thought principles into a living algorithm.
DeepSeek: Chinese Philosophy Meets AI Metrics
Feel free to leave a comment.
I have woven tales for anyone who cares to read them. My books await you on Google Books. You can also check out my stories on Medium.com.
I am eager to participate in research and produce content on Cross-Cultural Philosophy. Considering the many philosophy professors following Learn Vedanta Substack from universities across the five continents, I would be truly honoured to be involved in projects, as I have been recently approached. Please feel free to contact me.
I would be honoured if you considered subscribing to the Premium Contents of my Vedanta Substack and leaving feedback, comments, and suggestions on this page and by writing to me at cosmicdancerpodcast@gmail.com.