The Landscape of AI Definitions

The Landscape of AI Definitions

- by Irina Steenbeek, Expert in Artificial Intelligence

Discussions about artificial intelligence (AI) are ubiquitous today. I delved into over ten legislative documents from various world regions. The conclusion is eye-opening: AI lacks a universally agreed-upon definition.

The Landscape of AI Definitions

Countries approach AI definitions in three distinct ways:

1.    Formal Single Definition

The European Union uses a comprehensive, unified definition in its Artificial Intelligence Act, ensuring consistency across member states. Similarly, Canada adopts a single formal definition under its Artificial Intelligence and Data Act (AIDA) to clarify AI’s regulatory framework nationally.

2.   Formal Multiple Definitions

The United States employs multiple formal definitions, reflecting its diverse industries and federal structure. While this flexibility accommodates sector-specific needs, it creates inconsistencies in governance nationwide. The U.S. highlights the need for more coordinated AI regulation.

3.   No Formalized Definitions

Countries like Japan, Australia, the United Kingdom, Saudi Arabia, Brazil, China, and Singapore lack formal AI definitions. Instead, they rely on ethical guidelines and sector-specific principles to encourage responsible AI practices. While this approach provides adaptability, it risks legal interpretation and enforcement ambiguities.

Key Attributes of an AI System

Legislative definitions often describe AI as a “system” or “tools, technologies, and models.” Per DAMA-DMBOK2, a system is a unified whole of interdependent components, making “tools, technologies, and models” integral parts of such systems. Thus, AI can be defined as a system comprising various components, including data, tools, technologies, and models.

The core characteristics of an AI system are:

  • Autonomy: AI operates with varying levels of independence, making decisions without direct human involvement.
  • Processing: It processes input data—human or machine-generated—and abstracts it into models for automated analysis and inferences.
  • Adaptability: AI systems can learn and adapt post-deployment, improving effectiveness over time.
  • Outputs: They deliver outputs such as predictions, recommendations, decisions, or content, driving goal-oriented tasks.

Proposed AI Definitions

From these insights, the following definition encapsulates the essence of AI:

Artificial intelligence is a system that autonomously performs tasks using machine learning, data processing, and algorithmic models. It can adapt, learn, and improve based on data to achieve specific objectives such as prediction, classification, or optimization.

This definition aligns with the global regulatory discourse while clarifying AI’s multifaceted nature. As nations continue shaping their AI governance, establishing universally recognized definitions will foster innovation and collaboration worldwide.

Conclusion

The fragmented landscape of AI definitions underscores the need for harmonization in global governance. While each country’s approach reflects unique priorities, the push for clarity and consistency will only grow as AI transforms our world.