Defining Artificial Intelligence: Regulatory and Foundational Perspectives
The Need for a Comprehensive Definition for AI
Artificial Intelligence (AI) is at the forefront of technological innovation today, with increasing integration into various sectors, including healthcare and and biomedicine. This integration necessitates clear, comprehensive definitions of AI to guide development, regulation, and implementation. Here we examine the current FDA definition of AI, propose a foundational alternative definition, and discuss the importance of appropriate definitions in various contexts.
FDA's Definition of AI
The U.S. Food and Drug Administration (FDA) defines artificial intelligence as:
A machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions influencing real or virtual environments. Artificial intelligence systems use machine- and human-based inputs to perceive real and virtual environments; abstract such perceptions into models through analysis in an automated manner; and use model inference to formulate options for information or action.
Additionally, the FDA distinguishes Machine Learning (ML) as:
A set of techniques that can be used to train AI algorithms to improve performance at a task based on data.
The FDA definition demonstrates several strengths. It covers key aspects of AI systems, including their ability to make predictions, recommendations, and decisions. The definition acknowledges AI's influence on both real and virtual environments, encompassing a wide range of applications. It also outlines the AI process from input to output, including perception, abstraction, and inference. Importantly, it recognizes the role of human-defined objectives and human-based inputs, emphasizing the collaborative nature of AI systems.
However, the FDA's definition also has certain limitations that merit examination. It lacks explicit mention of AI systems' capacity for adaptation and improvement over time, a key characteristic of many contemporary AI technologies. The emphasis on "human-based" inputs may not fully account for the increasing autonomy and self-directed capabilities of advanced AI systems. The interdisciplinary nature of modern AI, which often incorporates insights from neuroscience, psychology, and cognitive science, may not be adequately reflected. Additionally, the definition does not explicitly address the scalability of AI systems or their potential to function as distributed, interconnected networks.
Foundational Definition of AI
A broader, more foundational definition of AI can be articulated as:
Artificial Intelligence refers to computational systems designed to process information, make decisions, and solve problems in ways that can match or exceed the capabilities of biological intelligence. These systems can analyze data, identify patterns, adapt based on input, learn from experiences, and perform tasks with varying degrees of autonomy.
This foundational definition offers several advantages. It broadens the scope of AI beyond human-like intelligence, acknowledging that AI can and does operate in ways that may be fundamentally different from biological intelligence. It provides a more neutral starting point for discussing AI, reducing the risk of anthropomorphizing AI or limiting our understanding of its capabilities and potential. The definition reinforces the idea that AI is fundamentally about information processing, decision-making, and problem-solving, regardless of how these capabilities are applied in specific contexts.
Furthermore, this definition allows for a more nuanced discussion of AI's potential. Instead of setting human-level performance as the benchmark, we can more easily discuss AI systems that operate at subhuman levels in some areas and superhuman levels in others. By removing the focus on simulating human intelligence, this definition encourages us to think more broadly about the nature of intelligence itself, inviting consideration of forms of intelligence that may be entirely alien to our current understanding.
Importance of Appropriate Definitions
Having appropriate and comprehensive definitions of AI is crucial for several reasons. Regulatory clarity is paramount, as clear definitions provide a framework for creating and enforcing regulations, helping determine which technologies fall under AI regulation and which don't. Consensus on appropriate definitions also promotes standardization across the industry, ensuring consistency in how AI is developed, implemented, and evaluated.
From a risk assessment perspective, accurate definitions help in identifying and assessing potential risks associated with AI technologies. This is particularly important in healthcare, where AI systems may be involved in critical decision-making processes that directly impact patient care. Public understanding is another key factor, as clear definitions aid in public education about AI, helping to dispel myths and set realistic expectations about AI capabilities and limitations.
In the legal sense, precise definitions are crucial for interpretation and decision-making in cases of liability or disputes involving AI systems. As AI becomes more autonomous, questions of responsibility and liability become increasingly complex, and clear definitions can provide a foundation for addressing these issues. For research and development, well-crafted definitions can guide efforts, ensuring that new technologies align with regulatory expectations.
As AI technologies continue to advance and permeate various sectors, the importance of clear, comprehensive, and adaptable definitions cannot be overstated. These definitions, whether regulatory or foundational, serve as the cornerstone for effective regulation, development, and deployment of AI systems. Regular reviews and updates of these definitions, informed by the latest technological realities and interdisciplinary insights, may be essential to ensure that our understanding and governance of AI remains relevant and effective in this rapidly evolving field.