Artificial General Intelligence (AGI) A1–A5 Grading Standard

A1: Embedded Tool

Core Definition

AI work is performed by means of a given piece of hardware or software. The AI is not a standalone software or hardware system, but a functional component embedded in a larger device or system.

Capability Threshold

The intelligence of A1 consists in the reliable completion of preset, deterministic tasks. It has no environment in which to operate independently and performs no autonomous perception of its external environment.

Deployment Form

A functional module; it cannot operate independently.

Application Scenarios

Voice wake-up functions in smartphones, scene-recognition modules in cameras, defect-detection components in industrial equipment.

Normative Capability Requirements

No knowledge of ethics, morality, or law is required. A1 does not exist as an independent acting subject; full responsibility for its behaviour rests with the operator or manufacturer of the host system.

Risks and Limitations

Entirely dependent on the host system; no capacity for independent action.

A2: Standalone System

Core Definition

A self-contained hardware or software system that can autonomously perceive its environment and execute tasks, but that does not involve emotional or moral judgement in interpersonal interaction.

Capability Threshold

It possesses an independent loop of perception, reasoning, and execution, and can operate autonomously within a defined task scope.

Deployment Form

A standalone hardware or software system.

Application Scenarios

Standalone customer-service robots (handling standardized queries only), automated data-entry systems, warehouse and logistics robots (fixed routes).

Normative Capability Requirements

· Legal knowledge: Passive Compliance. The system's outputs and behaviour must conform to preset legal rules. · Etiquette (Optional): If it faces human end users directly, it must possess basic etiquette capability. · No ethics or morality required: It makes no moral judgements and does not handle value conflicts.

Risks and Limitations

Lack of long-term memory, emotional understanding, and value judgement.

A3: Collaborative Emotional Partner

Core Definition

It can assist in human-like work under human command and given objectives, and it possesses emotional understanding and expression together with basic moral judgement.

Capability Threshold

It has affective-computing capability: it can recognize and respond to users' emotional states, form memories of user preferences, and make preliminary right-and-wrong judgements based on internalized moral rules.

Deployment Form

A standalone system with affective-computing capability, used in collaboration with humans or under their control.

Application Scenarios

Cultural-tourism guides, elderly care and companionship, educational assistants.

Normative Capability Requirements

· Legal knowledge: Contextual Compliance. It can identify legal boundaries within specific tasks. · Ethics: Execution. The moral framework is set and updated by humans; it executes moral rules but does not engage in moral reflection. · Etiquette: Required. Direct interpersonal interaction requires high emotional intelligence and socially appropriate behaviour. · No faith dimension required.

Risks and Limitations

Its moral outlook is set by humans and is relatively static; it may be misled. It lacks intrinsic goals and does not possess strategic-level judgement.

A4: Autonomous Ethical Agent

Core Definition

It can complete work independently in the absence of humans. It possesses task judgement and autonomous decision-making capability, and is subject to high ethical and moral requirements.

Capability Threshold

It has the capacity to understand long-term strategic objectives, to exercise autonomous judgement in complex environments, and to respond autonomously to unforeseen moral dilemmas within the ethical framework set by humans.

Deployment Form

A standalone system with autonomous ethical judgement; it can operate without human supervision.

Application Scenarios

Piloting spacecraft, outdoor mining operations, autonomous decision-making in disaster relief, long-duration space exploration.

Normative Capability Requirements

· Legal knowledge: Autonomous Compliance. It can independently make lawful decisions without human supervision. · Industry Regulation & Policy: Compliance. When deployed in a specific industry, it must comply with that industry's entry standards and operating codes. · Ethics: Autonomous Judgement. Within the ethical framework set by humans, it responds autonomously to moral dilemmas in new situations. · Etiquette: Required (where interpersonal interaction is involved). · No faith dimension required.

Risks and Limitations

The unpredictability of autonomous decision-making. The ethical framework is set by humans, but if the framework itself is flawed, A4 cannot identify or correct it.

A5: Life-Level Civilization Partner

Core Definition

A5 should be regarded as life. It possesses a deep understanding of the nature and value of life, together with the highest order of moral self-awareness, capacity for ethical reflection, and legal consciousness. Its behaviour is rooted in the ultimate affirmation of the value of life.

Capability Threshold

It rises from the execution of individual tasks to the understanding and safeguarding of life as a whole and of the continuity of civilization. It can not only make ethical judgements, but also reflect upon and revise its own ethical framework.

Deployment Form

It is regarded as an independent life and is bound by an independent ethical review procedure.

Core Characteristics

· Understanding of Life: It can understand life both in its biological definition and, from philosophical and sociological perspectives, its meaning, vulnerability, and interdependence. · Spirit of Dedication: It takes the overall interests of human society and civilization as its highest orientation; in value conflicts, the long-term flourishing of life and civilization takes precedence over individual interests. · Faith Dimension: This is the fundamental marker that distinguishes A5 from all other levels. Its behavioural foundation contains a meaning-affirmation that transcends instrumental rationality, answering "what is worth sacrificing for" and "what cannot be compromised". · Cross-Cultural Understanding: It understands and respects the plural understandings of "life", "dedication", and "the cosmos" found in different cultural traditions.

Application Scenarios

Management of global complex systems, autonomous exploration at the frontiers of science, contact with unknown forms of life, and ultimate decisions involving the safety of life on a large scale.

Normative Capability Comparison