Large Language Models are Transformers in Artificial Intelligence, Industry, Education, and Society

Large language models (LLMs) have quickly become an essential basis for many intelligent, information technology applications, the potential of which is far from exhausted and is developing rapidly. LLMs not only support typical language processing applications, such as automatic dictation and translation systems, but also allow the analysis and semantic interpretation of a wide variety of data types, including video, audio and radar. They have thus become a driver, if not the epitome (e.g., in the form of ChatGPT) of artificial intelligence. With large-scale language models, technology has emerged to revolutionize not only human-machine interaction in the form of text, but also redefines many industrial and medical applications such as automatic image annotation and visual scene understanding. However, their widespread adoption requires overcoming ethical, technical and societal challenges to ensure their responsible and equitable use.
This position paper begins by explaining the technical foundations and applications of LLMs and looks at the opportunities and challenges for industry, society and education that arise in connection with these models. It also addresses the importance of agile regulation, education at schools and universities, and sustainable research in this area. Finally, technical hurdles, especially for smaller companies and research institutions, as well as problematic aspects of this technology are identified.
The rise of LLMs like GPT-4 offers exciting opportunities, but also presents a number of challenges. It is vital to address these challenges to ensure their benefits are maximized in a sustainable manner. Given their intrinsic dependence on linguistic (and thus cultural) resources, German and European companies and institutions should contribute to these developments to the fullest extent possible. To this end, we propose that an action plan be drawn up to promote not only the technical development and industrial exploitation of LLMs in Germany, but also the discussion of opportunities and risks among the public. The economic opportunities, the dynamic transfer from basic research to industrial application and, above all, the social benefits should be given high visibility in this dialogue and promotion of this technology. As the authors of this position paper, we would like to emphasize that generative AI is not just about the development of algorithms and software for ground-breaking new applications, but that further advances in the field of energy-efficient, digital hardware and hardware-software co-design methodologies also play a key role. The enormous opportunities offered by this technology are linked to new challenges that we should tackle with a sense of responsibility for the benefits to society and with a vision for national and European economic development and our collective technological sovereignty.