Artificial Intelligence is broadly defined as the simulation of human intelligence by machines programmed to learn, reason, and solve problems [18]. AI encompasses several subfields, including machine learning (ML), deep learning (DL), natural language processing (NLP), and computer vision, each with distinct applications and limitations. AI can be categorized into two main types: 1) Narrow AI (Weak AI): AI systems designed to perform specific tasks, such as facial recognition, speech processing, or recommendation algorithms. These systems operate within predefined parameters and cannot generalize beyond their designated functions [20]. 2) General AI (Strong AI): Hypothetical AI that possesses human-like cognitive abilities, including reasoning, creativity, and self-awareness [9]. Unlike Narrow AI, General AI remains a theoretical construct with no existing examples in the real world [20].

Ismail Abbas M. Abbas , Thank you for this question! Here is a snippet from my article; Not sure if this fully answers your question - but it is indeed a thoughtful one! Please take a look at the full article: Article Artificial Intelligence: Myths and Facts

Artificial Intelligence (AI) can be simply defined as solving problems through automated algorithms Fig.2.

Automated programs can only be executed via statistical transition chains such as Markov chains and Cairo techniques statistical chains.

Excellent question Ismail. There are many possible definitions, but perhaps none of them are really satisfactory for all contexts. Nevertheless, it is important to have one or several good definitions that do not succumb to the vernacular use of the term "AI", which has degenerated to refer to any system running a computer program of any type. The word "software" no longer seems to occur in popular everyday communication.

The possibly first use of the term "artificial intelligence" dates back to the Dartmouth workshop of 1956. One of the organizers is credited with formulating AI to be “the science and engineering of making intelligent machines”.

A more recent academic definition in the textbook Artificial Intelligence: A Modern Approach, by Stuart Russell and Peter Norvig (2021), structures the field into system behavior objectives being either rational algorithmic or aiming to imitate human behavior. Furthermore, they differentiate between systems that simply compute ("think") or those that physically interact with the environment.

In the context of politics and regulations, definitions have been set forth by the OECD and by the European Union.

The OECD definition:

“An AI system is a machine-based system that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments. Different AI systems vary in their levels of autonomy and adaptiveness after deployment.”

The EU definition in the recently enacted AI Regulation 2024/1689/EU reads:

‘AI system’ means a machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptivenessafter deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments.

In both cases, the reader is left with the undefined terms "autonomy" and "adaptiveness" and will intuit his best understanding of these terms. Thus, these definitions might not be sufficiently stringent.

Perhaps I might share definitions for "automation", "autonomy", "adaptability", and "AI" that I have been motivated to set forth, coming from a background of safety of machinery and robotics, fields in which "AI" is already playing a strong role.

Automation:

The property of a technical system that expresses its capability to pursue its intended objectives without human assistance in the cases that have been explicitly anticipated at design time.

Confronting the system with unforeseen situations thus leads to the need for corrective intervention by a human operator.

The future state of an automated system operating in an intended manner can, in principle, be predicted.

Autonomy:

The property of a technical system that expresses its capability to continue pursuing its intended objectives without human assistance, even in case that it encounters situations or conditions not explicitly anticipated at design time.

Lack of such autonomy would be characterized by the need for corrective intervention by a human operator.

The future state of an autonomous system operating in an intended manner cannot, in principle, be predicted.

Degrees of autonomy can be defined for specific application areas but cannot be generalized to a generic scheme.

Adaptability (which I prefer over "adaptiveness"):

The capability of a technical system to modify its behavior appropriately as a response to a change in its environment or operating conditions.

Such modification can be necessary to allow the continuation of operation or to improve the pursuit of the system's objective(s).

An algorithm can be classified as “AI” if all ofthe following apply:

1.The algorithm has a well-defined objective.

2.The algorithm is designed to pursue the objective, either through providing output data that can be used for this by others (humans or algorithms) or through employing the modalities that the embodied algorithm has access to.

3.The proper functioning of the algorithm depends essentially on “experience” gathered in context, in the form of data describing the deployment environment. Such data can be provided from separate determinations (measurements, calculations, simulations, …) conducted in advance, before deployment of the algorithm. Such data can also be provided through online determinations conducted by the algorithm after deployment, using its embodiment (interfaces, sensors, actuators, …) and concordant access to relevant data.

4.The quality with which the algorithm pursues / fulfills its given objective improves with experience, in general.

Some comments on this definition:

Re. 1: As a rule, the objective is traceable to an original, human-specified objective. The algorithm is not, therefore, understood to be “autonomous” in the philosophical sense.

Re. 2: Here, the “embodied algorithm” can be taken to mean, for example, a robot with its control system, in which a running algorithm not only can give data as output but also can measure physical attributes of the environment and move a mechanism as a result ofits deliberations.

Re. 3: Without data, the algorithm is not useful. In one case, the dataset can be collected in advance and then “frozen” at deployment, which prevents the algorithm from changing its behavior after deployment. It will not “learn” in operation. In the other case, the dataset can be extended after deployment, which allows the algorithm to change its behavior after deployment, thus “learning” from experience.

Re. 4: As a rule, the whole reason that we go to the effort of putting in place data-driven methods is to allow for improvements of behavior with experience gathered.

Finally, I believe that there is no generally applicable definition. The field is so broad that actors in it will always need to develop a sharper definition for their own purposes in context. Nevertheless, perhaps the above can inspire you to find your own most useful definition(s).

Cheers,

Björn

If you don't understand artificial intelligence, ask yourself if Iam right because others don't understand it either.

There is no generally applic If you don't understand artificial intelligence, ask yourself if I'm right because others don't understand it either.able definition of artificial intelligence.

The field is so vast that its actors will always have to develop a more precise definition of their own objectives in their context.

Artificial intelligence is broadly defined as the simulation of human intelligence by machines programmed to learn, reason, and solve problems.

Therefore, AI encompasses several subfields, including machine learning (ML), deep learning (DL), robots, natural language processing (NLP), and computer vision, etc., each with distinct applications and limitations.

And yet, AI can be classified into two main categories:

i-Narrowly defined AI (concrete definition of AI)

Which is simply defined as an automated algorithm operating on the basis of statistical transition matrices, which allows it to increase the amount of existing information.

ii- General AI (broad definition of AI):

such as machine learning (ML), deep AI (DAI), image and voice recognition, etc.

These broad AI systems that operate according to predefined parameters and lots of input data cannot be generalized beyond their designated functions.

We believe that the term AI should be reserved for software capable of increasing the amount of information that is the function of only humans (and much less other creatures) and unitary 4D x-t statistical transition matrices.

Note that the link between artificial intelligence and 4D x-t unit space is revealed for the very first time.

Some hypothetical AIs with human-like cognitive abilities including reasoning, writing and reading rules, pronunciation, specific tasks such as facial recognition, speech processing, etc. are usually called AI systems, which is not true.

These systems operate in classic 3D+t space and obviously cannot increase the amount of existing information and therefore cannot be qualified as artificial intelligence.

We can conveniently call these systems storage intelligence systems (SI systems) rather than artificial intelligence systems.

On the other hand, concrete and narrow AI in 4D x-t unit space remains a well-defined theoretical and practical construct with many existing examples in physics and mathematics as well as in real daily life.