Since its release three years ago, ChatGPT has spread rapidly in Germany. People in Germany are commonly described as reserved and sceptical about AI. However, the figures show – even in international comparison – that Germany is not doing badly and that large language models are increasingly being used.
The application is mainly used in the private sphere and by young people. Companies, on the other hand, have so far tended to experiment with the technology rather than use it strategically. The application is particularly widespread in the information and communication industry, in knowledge-intensive services and in innovation-oriented industrial companies. The most common areas of application are text generation, customer support, marketing and software development. Furthermore, larger companies with more than 250 employees and a high degree of digitalisation rely on ChatGPT significantly more often than smaller companies.
This reluctance is partly due to the technical characteristics of the large language model. Specific resources such as data, computing capacity and energy are required to use AI. Furthermore, there are concerns because the technology systematically hallucinates, i.e. invents information. In addition, language models are based on digital data and are therefore inherently selective, because such data only represent a limited and artificially generated excerpt of reality.
On the other hand, however, there is criticism of aspects that are not related to the technology itself, but rather to the ChatGPT product. ChatGPT is criticised for its lack of transparency, reliability, specifications and pricing.
For Europe, therefore, the task is not simply to look at technical performance data and replicate ChatGPT with a delay. Rather, it is about advancing language models that are better in terms of their product characteristics. The opportunity for growth lies in adapting non-European models or developing our own models that are more transparent and reliable and better meet the requirements of specific contexts. This can also reduce energy consumption.
To achieve this goal, the right course must be set in business, politics and society. The focus must be on developing language models more in line with the requirements of specific sectors. This requires the development of skills, the harmonisation of sectoral rules and horizontal digital laws, access to resources and public data, and fair competition.
Read the entire analysis: “Drei Jahre ChatGPT – Wo stehen wir und welche Zukunft hat die europäische Wirtschaft?” here as PDF. Please note, to date the publication is only available in German.
