General News

Challenge: Legal definition of "artificial intelligence"

One of the greatest challenges for the legislative and advisory institutions is the definition of "artificial intelligence" (AI). 

If one looks at the proposed definitions of the European Commission and the European Parliament as well as the explanatory definitions, there seems to be more or less agreement on some aspects, while essential points seem to be open on others:

As things stand, there seems to be widespread agreement that the AI system (ultimately by humans) is given one or more goals to achieve. In this respect, however, it does not yet differ from other algorithms. 

In part, the proposed definitions also list (regulatory extensible) techniques and concepts that must be used by an AI system in order to qualify as an AI system. Whether an explicit list of techniques and concepts that are used to qualify as AI is useful or counterproductive, and which ones should be listed in what detail, is still a major point of discussion. 

The type of system is partly based on software, partly on a software-based system in a virtual environment or a system embedded in hardware. Against the background of research - such as "neuromorphic computers" oriented towards the human brain, which are not hardware and software in the conventional sense, biohybrid robots in which, for example, natural organisms are remotely controlled via electrical stimulation, or machines into which biological material is fed - a definition that is as broad as possible already seems to be purposeful.

The biggest question that remains however, is the definition of the criteria that make a system "artificially intelligent" when it uses an algorithm oriented to a given goal or when it uses certain concepts and techniques. Is it autonomy, rationality in the sense of a "best outcome", the learning element, does the purpose of the regulation matter, etc.? 

If one considers the large bracket of developments and fields of research that are taking place in this regard, it becomes clear why a definition must be a particular challenge. Ultimately, attempts are being made to increasingly technologise, understand, improve, facilitate, etc. all cognitive abilities (in the broadest sense) of humans but also other living beings with the help of various technologies or disciplines and concepts. And based on this, the question then arises: "Which of these should now be (specifically) regulated and why? 

So we can eagerly await what the final political agreement on such a complex development will look like in legal terms.