In terms of their frequency of use, the terms “artificial intelligence” (AI) and “machine learning” (ML) have now reached the status of buzzwords – and this frequency is reflected in the number of interpretations. Accordingly, it is worth establishing a common understanding:
- Artificial intelligence (AI): AI refers to the intelligence of non-living objects – that is, machines that are equipped with a broad-reaching analytic powers. This power gives rise to various – typically human – abilities such as perception, learning ability, reasoning, planning and the derivation of decisions. Depending on whether these abilities can only function in a particular context or can also be adapted to new contexts, a differentiation is made between weak and strong AI systems.
- Machine learning (ML): Within the framework of ML, computers are equipped to learn without first being explicitly programmed with specific details. This means that a computer will automatically recognise patterns and laws in the data provided. Upon completion of the learning phase, the experience gained is generalised – that is, converted into knowledge that can then be applied to new data sets by way of transfer. Accordingly, ML is essential for the realisation of artificial intelligence, since in essence, intelligence is nothing but learning. Just as people learn to communicate, recognise certain patterns (for example, in the form of grammar), or take into account rules when driving, machines can be trained to take over the related tasks independently. Since AI systems are generally based on ML, we often hear the two terms being used synonymously. For our purposes, we understand ML as a necessary prerequisite for AI.
The principle of AI is nothing new and the first theoretical concepts are already several decades old. However, it is only now that the technical prerequisites (in terms of sufficient hardware performance) exist for its potential to be exploited on an industrial scale and thus used to achieve improvements in performance, especially through automation. As shown in Figure 1, artificial intelligence solutions often consist of a combination of big data, artificial intelligence-based algorithms and the necessary high-performance hardware.
Figure 1: Artificial Intelligence solutions as the interplay of three central components
The possibility of transferring AI principles to different contexts results in an almost infinite variety of applications. Today, well-trained AI systems are able not only to perform certain tasks at a human level, but also to carry them out more quickly and with better endurance. In specific application fields, the ability of computers to handle complex data has even led to machine performance surpassing human performance. An example of this is the enhanced recognition of skin cancer by computers.
The high expectations placed on the potential of AI in the industrial environment are based mainly on the fact that the Internet of Things (IoT) provides the ideal prerequisites for AI applications due to the increasing prevalence of interlinked objects. The high data volumes generated within the framework of the IoT can be combined with AI to form specific value-creating applications such as predictive maintenance.
The impressive performance of these AI applications – not least thanks to the contribution of numerous Hollywood films – has stimulates our collective imagination in regard to the associated opportunities and risks. A scenario in which intelligent robots will make people completely obsolete will probably remain fiction for an indefinite period. In terms of cooperation between man and machine, AI serves as a supporting enabler for sustainable increases in performance. Accordingly, the processes and roles of people will change, but not their fundamental relevance. The following figure shows three examples of AI applications that can support people in specific ways in everyday life:
Figure 2: Examples of applications of artificial intelligence
Despite the associated potential, AI is definitely not a panacea for any and every business challenge. Our experience shows that specific application benefits must be systematically derived. The procedure shown in Figure 3 represents one way of doing this:
Figure 3: Procedural framework for the identification and integration of artificial intelligence solutions
In the procedure illustrated in Figure 3, steps 1 and 5 are particularly important for the success of artificial intelligence solutions. The AI approach must be based on concrete challenges and problems. The general question “For what can AI be used?” is too vague and is not conducive to meaningful answers.
Equally as important is the integration of AI solutions into existing organisations. If man-machine cooperation is not accepted by the employees involved, processes and roles are not adapted accordingly and the smartest solution is doomed from the outset. Accordingly, building the necessary trust is one of the essential prerequisites for successful human-machine cooperation. In particular, the introduction of AI solutions can lead to cultural challenges due to the fact that the process occurring between the input and output of the AI solution is often not transparent to employees. Against this background, the focus must lie not on the pre-defined algorithms, but on the data used to “train” the AI machine. Experience has shown that people need a certain amount of time to adapt, and this must be allowed for prior to full performance capacity being reached. In light of this, it is advisable not only to tackle the issue of AI at an early stage, but also from an organisational/cultural point of view, so that the full potential of AI can be diffused throughout the company.
Our experts will be delighted to support you in structuring and analysing the potential of artificial intelligence solutions in your company.
Contact one of our experts
Markus Fost, MBA, is an expert in e-commerce, online business models and digital transformation, with broad experience in the fields of strategy, organisation, corporate finance and operational restructuring.Learn more