Gerhard Schreiber

The article focuses on the Icelandic theologian Magnús Eiríksson (1806–1881), whose extensive writings have hitherto been a blind spot in research. The life and authorship of this controversial and independent thinker from the Golden Age of Denmark are sketched out with constant attention to his dogmatic stance, and key themes and motifs of his thought are explored with the aim of theological-historical contextualization. Remarkable similarities are found with the interpretation of the history of Christianity as a process of alienation, later prominently represented by Adolf von Harnack, while Eiríksson also proves to be a forward-looking theologian in other respects, making an in-depth study of his work both rewarding and profitable.

This chapter deals with AI-based text generators from an ethical point of view. It begins with a brief overview of generative AI (section 2) and AI-based text generators in particular. This is followed by a compilation of relevant areas of applied ethics (section 3), especially information ethics and its relatives, and then a structured, detailed ethical discussion (section 4). Opportunities and risks of operation and use are addressed. The chapter concludes with a summary and outlook (section 5). Overall, numerous risks are identified, as well as opportunities such as supporting work. The transformative potential for an important part of the creative and scientific professions is obvious. It is clearly a second wave of digitization that is taking hold.

Large language models (LLMs) such as ChatGPT are capable of impressive textual analysis and interpretation. This paper explores their potential as a „hermeneutic tool“ in historical research. Using a Middle High German chronicle as an example, the paper shows how ChatGPT can accurately categorize unfamiliar texts and clarify ambiguous passages. However, it reaches its limits when it comes to more specific historical contexts. By analyzing the underlying technology, the paper shows how LLMs appear to understand texts while actually generating them based on the „mechanics“ of statistically trained models. In doing so, this study demonstrates that these models can be of great assistance in historical research, but that true interpretation of historical texts and events still requires human understanding.

Like any science that deals with written artefacts, theology is faced with extreme challenges from text-generating programs such as ChatGPT. This is particularly true for Christian theology, insofar as its subject is a religion that refers to a holy scripture as its founding document. In this paper, I will conduct a thought experiment to find out whether a religion that is based on a text produced by artificial intelligence is conceivable. The answer is initially positive, since the humanity of the person conveying the revelation appears irrelevant at first glance. However, this result must be put into perspective from a theological point of view, since the humanity of the Holy Scripture must be understood in analogy to the humanity of Jesus Christ. Thus, a hypothetical religion whose founding document is an artificially generated text must be rejected for strictly theological reasons.

AI tools such as ChatGPT are already available for use in scholarly text production, and it is likely that their use will increase even further in the near future. However, not all AI-generated texts are the same, and therefore we should not treat them analogously - even if they share the common feature of being notationally identical to an existing text: By means of a thought experiment, the paper shows that for this particular set of AI-generated texts, we need to distinguish between copied and hallucinated texts, and derives from this questions regarding novel AI-generated texts that affect our understanding of science and scientific textual work at their very core.

Currently, the subject of artificial intelligence is increasingly becoming an ongoing focus of research in literary studies, due to the rise of Large Language Models such as ChatGPT. Extensive scholarly work, institutional connections, and research projects are just beginning to emerge. This article provides an overview of the present state of the discussion on artificial intelligence regarding the subject of literature and takes stock of relevant research desiderata. How is literature written by AI currentlydefined?What research questions arise from the establishment of AI in the literary market? And how is authorship renegotiated through AI literature? Perspectives from research into writing processes that consider the influence of AI on creative writing and storytelling will constitute the primary focus of the essay.

Current language models and generators such as ChatGPT have several characteristics that make them difficult to analyse by researchers in disciplines such as Natural Language Processing. These technologies are not publicly available and must be used on external servers, which makes them dependent on external models. This makes it much more difficult to reproduce research results because there is no way to store specific versions of the model, examine or change the code, and compare the results over time. This creates a hurdle in the reproducibility of research results, which is essential for scientific progress and credibility. Another significant problem is the explainability of the results. In particular, neural networks, which serve as the basis for language generators, are known to be „black boxes“. Although we can observe what outputs they produce, interpreting the generation process and the underlying decisions is still extremely challenging. This limitation of explainability makes it particularly difficult to use such models in sensitive application areas. There are several possible approaches to address these challenges. Promoting reproducibility requires supporting open and comprehensible research practices. Here, language models could serve as open-source projects with a transparent training process to provide deeper insights into the opportunitiesand risks of this technology. In terms of explainability, researchers should seek methods to better understand the decision-making of neural networks and other complex models.This may involve developing explanatory techniques and exploring models that are more interpretable.

This article presents an annotated German translation of Magnús Eiríksson’s “My Work as an Author” (1875), which can be regarded as his literary testament. In the secondary literature, Eiríksson is primarily known as a critical contemporary of Kierkegaard in Copenhagen. Despite writing over 20 monographs and around 50 articles in Danish, Swedish, and Icelandic, Eiríksson’s work as a whole has received little scholarly attention. In this final article, written in Danish, Eiríksson summarizes the key ideas of his life’s work, which this translation makes accessible to a wider audience.