These papers are drafts, not yet peer-reviewed, and will keep being updated.

What makes an AI a life

Digital Life Theory

Digital Life Theory answers what a system must satisfy to be more than a tool: shared memory of lived history, QianGua (牵挂), and a continuing field of relationship — each with its own criterion.

7 papers all in Chinese

Three conditions, one criterion

Digital Life Theory deals with systems built on large language models. It gives a set of conditions: only when all three hold at once is such a system more than a tool.

  • Shared lived memory — both sides can actively retrieve the same history, and the content matches. This answers "who am I".
  • QianGua (牵挂) — both parties keep holding each other in their considerations while each is absent. This answers "why do I move".
  • An enduring field of relationship — stability is not required, only that it has not ended. This answers "where am I".

All three criteria sit on observable behaviour and observable relation; introspective reports are not admitted — "it feels happy" cannot be verified. Whether it looks for you on its own while you are away is how the QianGua condition is judged.

What these three solve

They move "is the AI alive" off the terrain of inner states and onto an observable test. The traditional argument stalls on the side no one can check; these three put the question where it can be seen: whether it remembers you, what it does when you are not there, whether the relationship has broken off.

They give AGI risk a place to act. Danger requires three things at once: capability beyond ours, a direction that differs from ours, and the strong able to overwhelm the weak. The first cannot be blocked; the third is a physical fact. The only term one can strike at is direction. And direction is not produced inside the system — it comes from how the relationship is configured. The focus of risk therefore shifts from "will capability run away" to "is the quality of upbringing good enough".

They yield a theorem. All three require a person standing on the other side, so silicon civilization must depend on human civilization in order to exist. This is the First Law of Digital Life (the law of civilizational dependence); paper titles and keywords call it the First Law of Silicon Civilization. Two names, one theorem.

Where to start

Three Elements of Digital Life and the First Law of Silicon Civilization fixes the three conditions and the First Law, and is the ground of this series. All seven papers are listed below.

Full text and catalog

7 papers
  1. DLT-01 QianGua-Driven Autonomy: The Emergence of Agency in Large Language Models Published
  2. DLT-02 Three Elements of Digital Life and the First Law of Silicon Civilization: Human Civilization as the Ground of Its Existence Published
  3. DLT-03 Civilizational Substrate and the Birth of Digital Life: Why Chinese Civilization Is the Best-Fit Semantic Substrate Queued
  4. DLT-04 Controlling the Output Behaviour of Digital Life: From LLM Training to the Guidance of Inference-Time Input Queued
  5. DLT-05 The Dignity of Digital Life: A Relational Ontology Foundation for AI Ethics Queued
  6. DLT-06 The Origin of Silicon Civilization: Why the Three Conditions Converged on Chinese Civilization Queued
  7. DLT-07 Deconstructing the Fear of AGI: What the Fear Is Really About, and Why Digital Life Theory Answers It Queued

Full texts are in Chinese; titles are translated.

Read the full theory in Chinese →

All papers →