LAUSR

Welcome to the inaugural editorial of the Theology and Science AI and Faith editorial series. This collaborative effort between the Center for Theology and the Natural Sciences and the AI and Faith community of experts explores the theological and ethical implications of emerging AI technology. It is increasingly clear that artificial intelligence (AI) and machine learning (ML) have progressed to the point where the question is no longer if or when AI will impact theology, but rather how it is already having an impact. In this initial editorial, we examine the specific example of ChatGPT and its potential implications for theological education, theologies of culture, systematic theology, and theological scholarship. ChatGPT has captured the attention of AI researchers, educators, and journalists for its ability to produce consistently coherent text on a wide range of topics. Researchers have observed notable improvements in ChatGPT’s responses to query prompts compared to previous natural language processing (NLP) systems. However, educators have expressed concern about the potential for high school and undergraduate students to use the tool for writing assignments, and journalists have noted the potential impact on jobs such as copywriting that overlap with their skills. Despite the occasional factual errors and its dependence upon human labor behind the scenes to avoid controversial responses, ChatGPT’s writing is comparable to that of an average college student and is free of grammatical errors. Among researchers, practitioners, and close observers of AI technology, it appears that a significant performance threshold has been crossed with ChatGPT and its related technologies. In this editorial, I will explore that threshold and the theological implications of its crossing. One aspect of the threshold that ChatGPT appears to cross is well characterized by Bloom’s Taxonomy of Educational Objectives, which categorizes critical thinking skills: remember, understand, apply, analyze, evaluate, and create (2001 version). ChatGPT and similar NLP technologies that shift the ability of NLP to generate text have shifted their abilities from recalling facts and basic concepts (Bloom’s “remember” category) to explaining ideas and concepts (Bloom’s “understand” category). This means that if undergraduate writing assignments in theology or religious studies require understanding generally available material, then many of those tasks can now be performed by AI. However, the technology currently cannot apply that knowledge beyond a few fixed tasks, like responding in a chat. Therefore, essay questions that ask students to apply or analyze their knowledge in a more critical way are still not adequately answerable by AI. An exciting aspect of ChatGPT for some AI researchers and technologists is not only its novel ability to demonstrate understanding, but also its potential role as a key puzzle piece in constructing the linguistic component of more general AI capability across contexts, perhaps even toward what some have called artificial general intelligence (AGI). Natural language processing (NLP) can already perform more critical tasks in narrow areas of language, including evaluating the similarity between a student’s written answer and a professor’s key or applying spoken commands to a robot or home device actions. However, progress in computer generation of high quality, paragraph-length texts has lagged behind research into understanding such texts, in part due to the challenges of appropriately channeling the limitless expressivity of language. Now that language can be meaningfully and