Narratives of artificial intelligence frame public perceptions and expectations, and have a performative role, potentially leading to increased attention and resource allocation, acceptance of AI, or resistance to the technology. However, research on AI narratives frequently produces generalized and decontextualized accounts. This paper argues for closer examination of the specific processes that shape AI narratives in particular contexts. To explore this, nine AI-related exhibitions held in German museums from 2022 to 2023 were analyzed. The study draws on interviews with curatorial (...) teams and exhibition materials to investigate what narratives of AI museums are formed and how those relate to the dominant AI narratives described in previous research. It is shown that curators are critical of dominant AI narratives and aim to deconstruct common myths and present more nuanced portrayals of AI. Yet, they also face challenges in balancing this critical stance with the need for creating engaging, accessible exhibitions. The analysis reveals that while some curatorial strategies successfully counter AI stereotypes, others fail or inadvertently reproduce mainstream imaginaries. Curators discover that avoiding problematic understandings of AI in the way they had initially intended is challenging or even impossible. In other cases, when elements of mainstream AI discourse resurface in museum exhibits despite the critical stance of the curators, they remain unexamined, or are regarded as insignificant and inconsequential. (shrink)

Released mid-November 2022, Galactica is a set of six large language models (LLMs) of different sizes (from 125 M to 120B parameters) designed by Meta AI to achieve the ultimate ambition of “a single neural network for powering scientific tasks”, according to its accompanying whitepaper. It aims to carry out knowledge-intensive tasks, such as publication summarization, information ordering and protein annotation. However, just a few days after the release, Meta had to pull back the demo due to the strong hallucinatory (...) tendencies or underwhelming performances of the model. This article aims to study, through a critical threefold argument, the potential impacts of LLMs once deployed in the scientific value chain. Our first argument is a technical one. By examining the technicity of Galactica, it is possible to explain the descripancies between its promotional corporate discourse and abysmal outputs. Second, by going back to debates in both computer science and computational philosophy on the automation of abduction, we argue from the epistemological front that LLMs indeed cannot produce strong abductions and, therefore, claims about the automation of hypothesis generation remains chambering. Finally, our third argument is a sociological one. By conceptualizing the scientific field through Nancy Katherine Hayles’ cognitive assemblage theory, we aim to outline the potential steering of science by LLMs, mainly through information ordering. The core of our argument rests on the assertion that excessive control on information risks contravening a certain serendipitous aspect inherent to scientific discoveries. (shrink)