Page - 119 - in Digital Entrepreneurship - Impact on Business and Society

2 Theoretical Background for Entrepreneurs 2.1 The Technological Aspect of AI Aunitarydefinitionof artificial intelligence (AI), and thuswhen it canbeattributed to amachine, has so far been omitted because it differs according to the point of view. Essentially, the approaches to definition differ in their determination of intelligence, how this intelligencemanifests itself and atwhat point amachine can be considered intelligent (cf. Legg andHunter 2007, pp. 7–8). Nevertheless, it is possible to distinguish two forms of AI that are generally different in the way these programs work. First, the symbolic AI is a learning program that focuses on symbols as operating ground for its learning progress. A rule-based thinkingpresented in the symbols, e.g. numbers, drives this form.The major advantage of this type of artificial intelligence is its practicability for the human users. Because its logic is quite easy to understand for a human, it can be useful as a support system.Thismayprovide auser or companywith an advantage over another non-user, if used correctly.Their potential often lies in the automation ofprocesses,whichareusuallyconsuming intenseamountsof timeandknowledge. On the other hand, the symbolicAI is a program that needs a lot of programming work done into it. This iswhy the symbolicAI gets quite expensive (cf. Lee et al. 2019, pp. 2–3). The other major form of intelligence is the so-called neural AI, which is enabled by the usage ofmachine learning. Thismethoduses algorithmof improved learning by using practice or sample data as learning material for gen- erating patterns that it can later rely upon. The goal of this method is that the AI should function likeahumanbrain, in the sense that evencomplexproblemscanbe solved, similar to neural work. Thus, neural AI is not needing actual training embracedbyhumans in the formof teachers. Its objective is to learn by itself from data sources it is given access to (cf. Lee et al. 2019, p. 4). Especially,MLand its subfield deep learning are known from this area, due to major breakthroughs of these research fields especially in the field of image and voice processing (cf. Skilton andHovsepian 2018, pp. 132–134). Thisfield is receivingmore andmore attention from researchers andcompanies, because it has thepotential to learn from rawdata. Conversely, thismeans that less humanwork has to be done in advance. While bothmethods or types of AI still face problems in terms of scalability and reasoning,much hope is put into the combination of both types. The combination creates the hope of solving tasks related to fundamental problems in the applica- bility of the technology. Examples of these challenges are insufficient data for operations or the solution of the black-box appearance ofAI to a human observer. While this research stream is still in its infancy, entrepreneurs should pay attention to future opportunities (cf. Lee et al. 2019, pp. 4–5). Beyond the basic understanding of howAI can be differentiated from a tech- nological view, it is important for the entrepreneur, to knowwhatAI cando for his or her business. An approach to understand the way AI works in the practical context is to understand it as a predictionmachine. This logic focuses on the very AI-Enhanced BusinessModels for Digital… 123