After the initial success of OpenAI’s ChatGPT in 2022, the increasing availability of commercial Large Language Models (LLMs) has interrupted many cultural domains from the arts to writing and communication. Religious practices are no exception to this transformation. In 2023, hundreds of German Protestants attended a sermon delivered by an AI chatbot in St. Paul’s Church (Grieshaber 2023). A young theologian from the University of Vienna used ChatGPT to create a full church service—including prayers, music, and the sermon—and then created an avatar to deliver the service on a screen put up above the church’s pulpit. In 2024, this time at St. Peter’s Chapel, the oldest Catholic church in Switzerland, a theologian teamed up with AI researchers to build an “AI Jesus system”—a lifelike avatar of Jesus powered by AI—to offer spiritual advice in a confessional booth (Keaten 2024). Muslims, too, have embarked on the exploration and experimentation with LLMs. Mustafa Akçaya, a Turkish software engineer, is one of them. Using customization features that are available to OpenAI users/customers, he built KuranGPT that can answer Islamic questions in reference to the Qur’an (accessible at https://chatgpt.com/g/g-EW71Mjrk2-kurangpt). The premise of KuranGPT is that it will be a way of learning about Islam directly from the Qur’an without the need to listen to, in his words, “so-called clergy.”
Drawing on my anthropologicall fieldwork on AI and religion in the US and Turkey, this paper explores the cultural logic of un/translatability of the Qur’an. My primary focus will be on the design and perception of Qur’an-focused AI chatbots such as KuranGPT in Turkey. In the paper, I ask: What does AI-ification do to the Qur’an and its logic of un/translatability? How do AI enthusiasts coalesce their belief in the accessibility of the Qur’an’s text with their belief in the effectiveness and efficiency of algorithms? In what ways do the ideas of transparency, explainability, and interpretability traffic between AI and the Qur’an?
The paper is divided into three parts. The first part analyzes the techno-theological context within which KuranGPT and similar Islamic AI chatbots have turned AI into something more than an abstract philosophical question in Islamic circles in Turkey. With the availability of publicly accessible LLMs, AI is now both a vehicle and site for exploration and experimentation with the Qur’an and other Islamic texts. I will highlight a cultural affinitiy between the ‘query-based conversation’ of AI chatbots and the rhetorical style of many modernist Islamic intellectuals. Some—not all—of these intellectuals tend to approach the Qur’an as a source text a Muslim looks into to find ‘answers’ to questions related to practices and attitudes (e.g., what does the Qur’an say about inheritance?, is veiling required according to the Qur’an?, what awaits Muslims in ‘heavenly gardens’?). This is strikingly similar to the language of the simulated conversation in AI chatbots.
The second part of the paper examines the perception of KuranGPT and similar Qur’an-focused AI chatbots in Turkey. I will focus on the response from the Turkish state’s Religious Affairs Directorate. A few days after KuranGPT was made available, the Religious Affairs Directorate scholars warned Turkish citizens that Islamic knowledge does not come only from the Qur’an, and is shaped by many different kinds of scholarships, including those related to hadith and fiqh (Dengiz 2024). To be clear, the Religious Affairs Directorate is not against AI. Indeed, digitalization—especially the integration of AI into the Directorate’s services—was the topic of the most recent ‘Religious Assembly’ (Din Şurası) that took place in November 2024, and brought together scholars, intellectuals, and bureaucrats to discuss high-level strategies for the Directorate’s future direction. One of the ‘Religious Assembly’ recommendations was to examine all available Islamic AI chatbots and build a new one (potentially a "DiyanetGPT") that includes “valid religious knowledge” and is attentive to ‘cultural sensitivities.” Even though this recommendation seems to privilege Islamic scholars’ interpretive authority, it embraces the logic of AI-ification of Islamic texts, the Qur’an or otherwise.
The paper ends with a part that reflects on the predicament of the Qur’an’s translatability in the age of AI. Talal Asad (2018) writes that “the primary language of secular reason that employs numbers requires inattention to the world from which it has been abstracted” (p. 97). LLMs’ rendering of language is not just statistical but also dependent on probabilistic reasoning. Then, what happens to the theology of the Qur’an’s language in an LLM? Will AI bring forth a new conception of the Qur’an’s miraculousness (ijaz)? A Qur’anic verse frequently invoked for the Qur’anic doctrine of miraculousness is: “Surely if mankind and jinn banded together to bring the like of this Quran, they would not bring the like thereof, even if they supported one another” (Qur’an 17:88). Where will AI fit between “mankind and jinn”? What if people using LLMs think that they could produce something ‘like the Qur’an’? While Muslims tend to have different opinions on these questions, it is now possible to suggest that the current predicament of the Qur’an now lies not only at the interface between the languages of God and humans, but also in the language of machines as well.
Asad, T. (2018). Secular translations: Nation-state, modern self, and calculative reason. New York: Columbia University Press.
Dengiz, R. (2024, June 25). Diyanet'ten yapay zeka tabanlı KuranGPT'ye eleştiri: 'Fıkıhsız, hadissiz İslam eksik kalır’. Euronews Türkçe. Retrieved from https://tr.euronews.com/2024/06/25/diyanetten-yapay-zeka-tabanli-kurangptye-elestiri-fikihsiz-hadissiz-islam-eksik-kalir
Grieshaber, K. (2023, June 10). “Can a chatbot preach a good sermon? Hundreds attend church service generated by ChatGPT to find out.” Associated Press https://apnews.com/article/germany-church-protestants-chatgpt-ai-sermon-651f21c24cfb47e3122e987a7263d348
Keaten, J. (2024, November 28). “‘AI Jesus’ avatar tests man’s faith in machines and the divine.” Associated Press https://apnews.com/article/artificial-intelligence-chatbot-jesus-lucerne-catholic-66268027fbcf4b48972d1d62541f0b16
Many Muslims have recently embarked on the exploration and experimentation with LLMs (Large Language Models) on religious texts. Some are developing their own AI chatbots based on the Qur'an. As these AI models are designed to respond to queries about Islam, AI enthusiasts coalesce their belief in the "transparency" of the Qur’an with their belief in the "transparency" of algorithms. Examining the design and perception of Qur’an-focused AI chatbots, this paper explores the cultural logic of un/translatability of the Qur’an in the age of LLMs: What does AI-ification do to the Qur’an and its un/translatability? In what ways do the ideas of transparency, explainability, and interpretability traffic between AI and the Qur’an? If LLMs’ "understanding" of language is not just statistical but also dependent on probabilistic reasoning, what happens to the theology of the Qur’an’s language? Can AI bring forth a new conception of the Qur’an’s miraculousness (ijaz)?