‘‘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)
An algorithm is converted into an AI system when it operates with varying degrees of autonomy and starts to generate significant outputs like predictions, recommendations, or decisions that can impact physical or virtual environments. This suggests that AI systems go beyond simple tasks and are designed to interpret complex data and make decisions without explicit human intervention, reflecting computational learning and an intuitive decision-making process.
’‘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)
This definition of a ‘foundation model’ adds further insight into what constitutes an AI system. Fundamentally, a foundation model, which is an AI system, is trained on diverse and large-scale data. It is designed to provide a wide range of outputs and can be adapted to perform multiple distinctive tasks. This highlights more advanced forms of AI, capable of generalizing knowledge across various domains rather than being narrowly bounded.
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)
The defining characteristics of an AI system –learning, reasoning, and modeling capabilities– differentiate it from simpler software or less complex programming methods. Thus, an algorithm would be recognized as AI when it embodies these properties. The purpose is to establish a clear definition that ensures legal certainty and can adapt to abrupt advancements in AI technologies.
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. (Recital 6)
AI is depicted as an entity with some level of autonomy, meaning that it can function somewhat independently from human control and intervention. This autonomy is what distinguishes AI systems from basic algorithms, making them AI once they acquire a self-sufficient operational capacity.
The developers of free and open-source AI components should not be mandated under this Regulation to comply with requirements targeting the AI value chain and, in particular, not towards the provider that has used that free and open-source AI component. (Recital 12c)
This recital underscores the treatment of free and open-source AI components under the Regulation. It does not define when an algorithm becomes AI directly, but it provides a perspective on how the law regards different AI components and their creators.
In conclusion, an algorithm is considered as AI under the EU AI Act when it manages to function with a certain level of autonomy, can perform a range of tasks, and has learning and reasoning abilities. It goes beyond the capabilities of a traditional algorithm, acquiring a type of self-sufficiency and decision-making potential that sets AI apart. Unfortunately, the information from Article 56b and Annex I was not accessible and could not be included in the analysis.
“‘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))
Article 3(1) of the EU AI Act postulates that an algorithm becomes an AI when it operates autonomously to achieve explicit or implicit goals, producing outputs such as predictions, recommendations, or decisions that influence physical or virtual environments. These outputs should result not strictly from the initial input data, but from the AI’s operations.
”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)
In Recital 6, the Act underlines that the definition of an AI system should align with international standards while leaving room to adjust to rapid advancements in AI technology. This is meant to keep the Act relevant in the face of evolving technology.
”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)
Furthermore, AI systems possess key characteristics like learning, reasoning, or modeling capabilities that distinguish them from simpler software systems or programming approaches. This suggests that an algorithm can be considered as AI when it exhibits these defined capabilities.
”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.” (Recital 6)
The statement identifies autonomy as a chief characteristic of AI systems. It implies that the ability to operate independently of human intervention can be a determining factor in classifying an algorithm as an AI system.
”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.” (Recital 6)
This observation points to the output generation capability of AI systems, which may include making predictions.
”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)
As per Recital 6a, machine learning in AI allows the system to adapt autonomously and perform new tasks that weren’t explicitly programmed. This optimization of models based on input data is part of the process wherein an algorithm becomes AI.
”The function and outputs of many of these AI systems are based on abstract mathematical relationships that are difficult for humans to understand, monitor and trace back to specific inputs. These complex and opaque characteristics (black box element) impact accountability and explainability.” (Recital 6a)
The complex relationships and the “black box” nature of these systems, whereby their functioning is challenging for humans to comprehend, implies the algorithm is AI.
”Comparably simpler techniques such as knowledge-based approaches, Bayesian estimation or decision-trees may also lead to legal gaps that need to be addressed by this Regulation, in particular when they are used in combination with machine learning approaches in hybrid systems.” (Recital 6a)
Even relatively simpler algorithmic solutions like knowledge-based approaches, Bayesian estimates, or decision trees can be seen as AI when combined with machine learning in hybrid 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.” (Recital 6b)
An algorithm becomes AI when it is incorporated into a functioning system, whether standalone, embedded in a physical product, or serving as a non-embedded component of a larger system that exhibits or engages in a form of artificial intelligence.
”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)
If an AI component is critical for the functioning of a larger system, then the entire larger system is considered as one single AI system under the EU AI Act.
In summary, the definition of AI in the EU AI Act is comprehensive and multifaceted. An algorithm becomes AI under the Act when it’s able to operate autonomously, generate outputs influencing environments, and adapt through learning. Plus, the application of an algorithm in a system—standalone or as a part of a hybrid system—also influences its designation as AI. However, it’s important to note that due to the dynamic nature of AI, maintaining legal certainty through updated interpretations is crucial.
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. (Recital 6)
This excerpt from Recital 6 highlights that AI systems are distinguished from other software systems based on their learning, reasoning, or modeling capabilities. This means that an algorithm becomes an AI when it can learn, reason, or model scenarios.
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. (Recital 6)
Another key quality described in Recital 6 is the level of autonomy. For an algorithm to qualify as AI, it should have some level of independence from human interaction, meaning it can operate without constant human monitoring or input.
The term “machine-based” refers to the fact that AI systems run on machines. (Recital 6)
Being machine-based is another distinguishing characteristic. This means that AI runs on hardware and is not a mere abstract concept or methodology.
The reference to explicit or implicit objectives underscores that AI systems can operate according to explicit human-defined objectives or to implicit objectives. (Recital 6)
This part reflects that AI systems should be able to operate according to either explicit objectives set by humans or implicit objectives inferred from the input data and learning process.
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. Machine learning approaches include, for instance, supervised, unsupervised, and reinforcement learning, using a variety of methods including deep learning with neural networks. (Recital 6a)
Further clarification provided in Recital 6a is that machine learning is a key attribute of AI. If an algorithm can learn and optimize from the input data it receives, adapting and performing new tasks autonomously, it fits into the definition of AI.
AI systems can evolve after deployment. (Recital 6a)
Another significant aspect is the ability of an AI system to evolve after its deployment. This ‘learning’ component is an essential part of the definition of AI.
Comparably simpler techniques such as knowledge-based approaches, Bayesian estimation or decision-trees may also lead to legal gaps that need to be addressed by this Regulation, in particular when they are used in combination with machine learning approaches in hybrid systems. (Recital 6a)
Lastly, even simpler methods such as knowledge-based approaches can qualify an algorithm as AI if they are used in combination with machine learning.
In conclusion, according to the EU AI Act, an algorithm becomes an AI when it has learning, reasoning, or modelling capabilities, operates autonomously to some extent, has explicit or implicit objectives, is capable of evolving after deployment, and can potentially combine simpler techniques with machine learning.
‘‘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))
Under Article 3(1), an algorithm becomes an artificial intelligence system when it is machine-based and designed to operate with varying degrees of autonomy. The key feature of an AI system per this definition is that it can generate outputs including but not limited to predictions, recommendations, or decisions. These outputs impact physical or virtual environments. Therefore, the autonomy and the ability to create meaningful outputs that can influence real or virtual environments are major defining factors of AI in this context.
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)
This quote defines an AI system as being distinguished from simpler software systems or programming approaches. The distinction is based on certain key characteristics of the AI, such as learning, reasoning, or modeling capabilities. This aligns with common understandings of AI as systems with capabilities beyond fixed, pre-programmed rules, instead having the ability to adapt and improve over time via learning.
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. (Recital 6)
This quote talks about another distinguishing feature of an AI system, its autonomy. An AI system operates with at least some degree of independence from human control, and can function without any human intervention.
AI systems often have machine learning capacities that allow them to adapt and perform new tasks autonomously. (Recital 6a)
Building on the previous definition, this quote describes the machine learning capacities of an AI system as one of its characteristics. Machine learning is considered as the adaptation capability of the AI system that lets it perform new tasks independently.
The function and outputs of many of these AI systems are based on abstract mathematical relationships that are difficult for humans to understand, monitor and trace back to specific inputs. (Recital 6a)
Finally, this quote emphasizes the inherent complexity of AI systems. The functions and outputs are often based on complicated mathematical relationships that can be difficult for humans to interpret, leading to an aspect of opaqueness or the so-called ‘black box’ element of AI. These recitals outline the distinguishing features of an AI system. Thus, an algorithm becomes AI when it demonstrates characteristics like learning, autonomy, ability to reason and adapt to new tasks.