Austrian Research and Technology Report 2020

3. Artificial Intelligence (AI)
161
3.1 Context
Digitalisation is a major trend in national and interna-
tional innovation systems (see also Austrian Re-
search and Technology Report 2019), and the area of 
artificial intelligence (AI) is attracting particular at-
tention. The increasing use of AI and its rapid tech-
nological development at present are benefiting 
most notably from the availability of large volumes of 
data (big data), the rapid growth in computers’ pro-
cessing power and algorithms that are constantly 
improving. 
Based on the European Commission’s definition of 
the term, 
artificial intelligence
means artificial sys-
tems that appear to demonstrate intelligent be-
haviour. These systems analyse their environment 
and act with a certain degree of autonomy in order 
to achieve specific objectives. They can be soft-
ware-only systems that perform actions in virtual en-
vironments or systems embedded in hardware, such 
as smart robots, drones and autonomous vehicles.
140
An important distinction is made in the AI field be-
tween narrow and general AI.
141
A general AI system 
is conceived as a system capable of carrying out 
most of the activities that humans can. Narrow AI 
140 See European Commission (2019a).
141 See Nilsson (2009).
142 See Prem and Ruhland (2019).
systems, by contrast, are able to perform one or a 
few specific tasks. The AI systems currently in use 
are all examples of narrow AI.
This definition of AI presents significant obstacles 
to more detailed analysis. In particular, research pa-
pers/findings, applications, companies and projects 
in this field are difficult to classify due to the 
duality 
of applications and technologies
(i.e. AI refers both 
to the technology/technologies used and to a broad 
range of different applications). Another issue pre-
venting clear categorisation is the simple fact that 
many AI disciplines currently employ methods bor-
rowed from other fields – robotics and speech recog-
nition often use learning systems, for instance, while 
many modern robots use image analysis systems.
142
As Fig. 3-1 clearly shows, AI covers much more 
than just learning systems that approximate func-
tions based on an extensive data pool. Modern AI 
research encompasses a large number of areas that 
do not use any data at all to create models or solve 
problems, such as “searching” or “planning”. However, 
the feasibility of an AI project is heavily dependent 
on the structure of the problem at hand and often 
requires in-depth expert knowledge of the relevant 
domain.