During my research for the book “AI Compass for Decision Makers” I kept coming across the question of what is actually meant by artificial intelligence and what I understand by it. The variety of definitions is indeed broad, and it is not always easy to distinguish an intelligently programmed software solution from a solution that is actually based on methods of artificial intelligence.

A somewhat official definition comes from the homepage of the EU Commission and was published in April 2019. Under the title “A Definition of AI: Main capabilities and disciplines“, however, the restriction is found again that although this definition was commissioned and published by the EU Commission, it should under no circumstances be treated as an official EU definition.

In June 2018, the Commission had set up a so-called High-Level Expert Group on Artificial Intelligence (AI HLEG). It ultimately consisted of 52 experts. Representatives of corporations such as Airbus, Bosch, IBM and SAP were among them, as well as numerous scientists from various universities and research institutions, with disciplines ranging from robotics and data science to political consulting and theoretical philosophy. Ethics specialists such as the representative of the aid association for the blind and weak and bankers were also among them. The definition of AI was the basis on which the group of experts then began to formulate guidelines for trustworthy AI.

In the “AI Compass for Decision Makers” I have reproduced the definition in the form of a translation of my own. In doing so, I limited myself to the short version which is recommended for use by the expert group at the end of the paper. Here is the original:

Artificial intelligence (AI) systems are software (and possibly also hardware) systems designed by humans that, given a complex goal, act in the physical or digital dimension by perceiving their environment through data acquisition, interpreting the collected structured or unstructured data, reasoning on the knowledge, or processing the information, derived from this data and deciding the best action(s) to take to achieve the given goal. AI systems can either use symbolic rules or learn a numeric model, and they can also adapt their behavior by analyzing how the environment is affected by their previous actions.

As a scientific discipline, AI includes several approaches and techniques, such as machine learning (of which deep learning and reinforcement learning are specific examples), machine reasoning (which includes planning, scheduling, knowledge representation and reasoning, search, and optimization), and robotics (which includes control, perception, sensors and actuators, as well as the integration of all other techniques into cyber-physical systems).

(KI-Kompass für Entscheider, Hanser Verlag, Munich, to be published in August 2020, chapter 2.1)

You search in vain for the word “autonomous” here. The elevation of AI to a technology that possibly replaces and renders superfluous human beings in their decision-making ability and ultimate responsibility was obviously not favored by the expert group. Instead, this definition focuses on what is actually possible with data, data science and machine learning.

The question of which decisions are delegated to AI by humans, which measures can and should be taken by AI in practice, is the subject of the Guidelines for Trustworthy AI, which were also published by AI HLEG in April 2019.