‘‘artificial intelligence system’ (AI system) means a machine-based system that is designed to operate with varying levels of autonomy and that can, for explicit or implicit objectives, generate outputs such as predictions, recommendations, or decisions, that influence physical or virtual environments; (Article 3)
According to Article 3, an artificial intelligence (AI) system is a machine-based system which can operate with different levels of autonomy, meaning it can independently function, make decisions, and generate outcomes to various degrees. These outputs are generated based on explicit or implicit objectives and can impact both physical and virtual environments, suggesting its application in a broad range of areas from recommending movies based on user’s viewing history, to self-driving cars making decisions on road.
The notion of AI system in this Regulation should be clearly defined and closely aligned with the work of international organisations working on artificial intelligence to ensure legal certainty, harmonization and wide acceptance, while providing the flexibility to accommodate the rapid technological developments in this field. Moreover, it should be based on key characteristics of artificial intelligence, such as its learning, reasoning or modelling capabilities, so as to distinguish it from simpler software systems or programming approaches. (Recital 6)
Recital 6 expands on this definition, stressing clarity and the need for distinguishing AI from conventional software systems. Notably, these are the learning, reasoning, and modeling capabilities of AI systems. It also emphasizes harmonization with the work of international organizations for wide acceptance and legal certainty, while allowing for flexibility to account for rapid technological advancements.
AI systems often have machine learning capacities that allow them to adapt and perform new tasks autonomously. Machine learning refers to the computational process of optimizing the parameters of a model from data, which is a mathematical construct generating an output based on input data. (Recital 6a)
Recital 6a further clarifies that AI systems often possess learning capabilities, allowing them to perform new tasks independently. This “machine learning” capacity, which optimizes model parameters based on input data, helps distinguish AI from traditional systems.
AI systems can be used as stand-alone software system, integrated into a physical product (embedded), used to serve the functionality of a physical product without being integrated therein (non-embedded) or used as an AI component of a larger system. If this larger system would not function without the AI component in question, then the entire larger system should be considered as one single AI system under this Regulation. (Recital 6b)
Lastly, Recital 6b outlines that AI can exist both as a standalone system or be embedded into another system or product. It emphasizes that a whole system in which an AI component is integral should be considered a singular AI system within the context of this Regulation. This provides clarity on when a system would fall under the scope of this Regulation.
In conclusion, the AI Act defines AI as a machine-based system with various degrees of autonomy, capable of influencing physical or virtual environments, and distinguished by its learning, reasoning, or modelling capabilities. It can exist in various forms and setups, and broader systems which are dependent on these capabilities are also considered to be AI systems.
“‘artificial intelligence system’ (AI system) means a machine-based system that is designed to operate with varying levels of autonomy and that can, for explicit or implicit objectives, generate outputs such as predictions, recommendations, or decisions, that influence physical or virtual environments;” (Article 3)
According to Article 3 of the proposed AI Act, an artificial intelligence (AI) system is defined as a machine-based system designed to operate with varying levels of autonomy. This means an AI system can operate or produce results independently, with less or more human intervention, depending on the given situation or the particular design of the system.
This system also has the ability to generate outputs such as predictions, recommendations, or decisions, meaning it can analyze data and produce results or course of action based on its analysis. This implies that AI systems are designed to aid decision-making or influence actions, possibly in different fields or applications where decisions-making or predictions are required.
Lastly, the AI system can influence physical or virtual environments. This indicates that the effects or outputs of AI systems can have real-world implications on physical environments, such as industrial automation, self-driving cars, etc., or affect virtual environments like recommendation systems in online shopping, personalized content in streaming platforms, or virtual assistants.
”The notion of AI system in this Regulation should be based on key characteristics of artificial intelligence, such as its learning, reasoning or modelling capabilities, so as to distinguish it from simpler software systems or programming approaches. AI systems are designed to operate with varying levels of autonomy, meaning that they have at least some degree of independence of actions from human controls and of capabilities to operate without human intervention. The term “machine-based” refers to the fact that AI systems run on machines.” (Recital 6)
This segment of Recital 6 defines an Artificial Intelligence (AI) system as distinguished from simpler software systems or programming approaches by its key characteristics. These characteristics include learning, reasoning or modeling capabilities, meaning AI systems can progress, reason, and simulate behavior or events. Also, AI systems are identified by their level of autonomy or their ability to function with at least some degree of independence from human control and can operate without any human intervention. In addition, the term “machine-based” implicitly implies that AI systems are run on machines, which further sets them apart from other software systems that are not “machine-based."
"The reference to explicit or implicit objectives underscores that AI systems can operate according to explicit human-defined objectives or to implicit objectives. The objectives of the AI system may be different from the intended purpose of the AI system in a specific context.” (Recital 6)
In the second extracted portion of Recital 6, AI systems are described as being driven by either explicit or implicit objectives. Explicit objectives might be human-defined goals which the AI system is programmed to meet. Implicit objectives are those that the AI system devises or interprets on its own, and which might not always align perfectly with the explicit human objectives. The context-dependent behavior of AI indicates its flexibility.
”The reference to predictions includes content, which is considered in this Regulation a form of prediction as one of the possible outputs produced by an AI system. For the purposes of this Regulation, environments should be understood as the contexts in which the AI systems operate, whereas outputs generated by the AI system, meaning predictions, recommendations or decisions, respond to the objectives of the system, on the basis of inputs from said environment. Such output further influences said environment, even by merely introducing new information to it.” (Recital 6)
In the third extract from Recital 6, the process of an AI system’s operation is explained. According to this paragraph, an AI system takes inputs from its operational context or environment, applies its learning, reasoning, and modeling capabilities to those inputs, and generates outputs. The outputs can be predictions, recommendations, or decisions that correspond to the AI system’s objectives. Furthermore, these outputs do not simply translate into outcomes but go on to influence the environment by introducing new information, showcasing the AI system’s capability to actively shape the context in which it operates.
“‘artificial intelligence system’ (AI system) means a machine-based system that is designed to operate with varying levels of autonomy and that can, for explicit or implicit objectives, generate outputs such as predictions, recommendations, or decisions, that influence physical or virtual environments.” (Article 3 - (1))
According to Article 3 of the AI Act, an Artificial Intelligence (AI) system refers to a machine-based system which can operate with different levels of autonomy. It could perform tasks that influence either a physical or a virtual environment, based on the objective for which it has been designed. This could include generating decisions, predictions, or recommendations.
The notion of AI system in this Regulation should be clearly defined and closely aligned with the work of international organisations working on artificial intelligence to ensure legal certainty, harmonization and wide acceptance, while providing the flexibility to accommodate the rapid technological developments in this field. (Recital 6)
Recital 6 emphasizes that the definition of an AI system provided in this regulation is based on international standards while maintaining flexibility to accommodate rapid technological advancements in this field. This ensures that the definition captures the broad scope of applications and is adaptable to cover future developments in AI technology.
AI systems often have machine learning capacities that allow them to adapt and perform new tasks autonomously. Machine learning refers to the computational process of optimizing the parameters of a model from data, which is a mathematical construct generating an output based on input data. (Recital 6a)
Recital 6a outlines machine learning as a pivotal aspect of AI systems, enabling them to adapt and perform tasks autonomously. The fundamental attribute of machine learning includes the computational process of optimizing parameters from input data to generate an output.
AI systems can be easily deployed in multiple sectors of the economy and society, including cross border, and circulate throughout the Union. (Recital 2)
Finally, Recital 2 highlights the broad and cross-sectoral application of AI systems, demonstrating the far-reaching influence of AI across different sectors and cross-border regions, further emphasizing the need for a comprehensive definition and corresponding regulations.
In summary, the definition of AI according to the EU AI Act covers machine-based systems operating with variable autonomy and capable of generating outputs that influence physical or virtual environments. These definitions consider the core characteristics of AI such as flexibility, adaptability, and broad applicability across economies and borders. Legal interpretations of the definition reflect efforts for international alignment, broad inclusivity, and preparation for future technological advancements in AI.
‘artificial intelligence system’ (AI system) means a machine-based system that is designed to operate with varying levels of autonomy and that can, for explicit or implicit objectives, generate outputs such as predictions, recommendations, or decisions, that influence physical or virtual environments; (Article 3(1))
According to the AI Act, an AI, or Artificial Intelligence system, is a machine-based system. The key characteristics of these systems include their ability to operate with various degrees of autonomy and generate outputs. These outputs can take the form of predictions, recommendations, or decisions that can make an impact on both physical and virtual environments. This means that an AI system could potentially affect the real world or digital spaces, such as the algorithms determining the content you see on social media.
‘foundation model’ means an AI system model that is trained on broad data at scale, is designed for generality of output, and can be adapted to a wide range of distinctive tasks; (Article 3(1c))
AI structures are often based on models referred to as ‘foundation models’. According to the AI Act, these models are versatile, designed to generate a wide variety of outputs and be adaptable to numerous distinctive tasks. These models are trained on a broad range of data at a large scale, suggesting extensive and comprehensive computational processing and learning.
‘general purpose AI system’ means an AI system that can be used in and adapted to a wide range of applications for which it was not intentionally and specifically designed; (Article 3(1d))
A ‘general purpose AI system’ is defined as a type of artificial intelligence system that can be adapted and used across a wide range of non-specific applications, even if those applications were not the ones it was intentionally designed for. This means that such a system has high adaptability and shows flexibility in what it can be used for.
The notion of AI system in this Regulation should be clearly defined and closely aligned with the work of international organisations working on artificial intelligence to ensure legal certainty, harmonization and wide acceptance, while providing the flexibility to accommodate the rapid technological developments in this field. (Recital 6)
This quote indicates that the European AI Act views AI, or Artificial Intelligence, as a concept to be properly defined and aligned with international organizational standards. It expresses the necessity for a clear definition that is legally certain, harmonized, and widely accepted, while also flexible enough to accommodate rapid technological developments in the field of AI.
Moreover, it should be based on key characteristics of artificial intelligence, such as its learning, reasoning or modelling capabilities, so as to distinguish it from simpler software systems or programming approaches. (Recital 6)
This section points out that AI, according to the European AI Act, should exhibit unique characteristics such as learning, reasoning, and modelling capabilities. This distinction is necessary to differentiate AI from simpler software systems and programming approaches, indicating that its capabilities extend beyond basic computational tasks.
‘human agency and oversight’ means that AI systems shall be developed and used as a tool that serves people, respects human dignity and personal autonomy, and that is functioning in a way that can be appropriately controlled and overseen by humans; (Article 4a)
This section emphasizes that AI systems are tools designed to serve human needs, and as such, they should respect human autonomy and dignity. Additionally, these systems will need to be operated and supervised by humans to ensure they are functioning as intended.
From analyzing the aforementioned “Article 3”, “Article 4a” and “Recital 6”, AI or Artificial Intelligence in the context of this regulation, is a system or tool to serve, assist, and benefit humans and society. AI systems should embody these principles: human agency and oversight, technical robustness and safety, privacy and data governance, transparency, diversity, non-discrimination and fairness, and social and environmental well-being.