Stanford
Love Data Week

AI & Astronomy Imagery

Further Reading

Here we have collected scholarly articles examining how generative AI could be used for diverse applications across a variety of scientific fields.  For instance, generative AI frameworks enhance multimedia education by creating situational videos for better comprehension in teaching Tang poetry (Chen & Wu, 2024) and help visualize complex physics concepts like relativity for high school students, aiding in their understanding and engagement (de Souza et al., 2024). However, generative AI technology also presents challenges, such as AI-enabled image fraud in scientific publications (Gu et al., 2022) and potential integrity issues in biomedical research (Zhu et al., 2024). Moreover, while AI can augment public discourse and engagement (Riling et al., 2024), existing models require refinement to address biases and inaccuracies when generating images about nuclear energy (Joynt et al., 2024). Similarly, in Earth sciences, generative AI tools show promise but need careful bias management (Kupferschmidt et al., 2024). These perspectives reflects a broader consensus that the implementation of generative AI across fields requires a balanced approach of thoughtful consideration of both its potential benefits and possible pitfalls.

  • Chen, X., & Wu, D. (2024). Automatic generation of multimedia teaching materials based on generative AI: Taking Tang poetry as an example. IEEE Transactions on Learning Technologies, 17, 1327-1340. https://doi.org/10.1109/TLT.2024.3378279

This article introduces a technical framework for generating Tang poetry situational videos using generative AI to create multimedia teaching resources, composed of modules for textual comprehension, image creation, and video generation. The framework incorporates various technologies and demonstrates that the generated videos and images enhance students’ understanding and reduce cognitive load, highlighting the potential of generative AI in education.

  • de Souza, M., Won, M., Treagust, D., & Serrano, A. (2024). Visualising relativity: Assessing high school students' understanding of complex physics concepts through AI-generated images. Physics Education, 59(2), 025018.

This article explores the use of generative AI to help students visualize and understand Einstein's relativity theory, analyzing students' prompts and AI-generated images to assess engagement and comprehension. The findings indicate that AI-generated visuals can serve as an effective educational resource, offering insights into students' conceptual grasp of RT and shedding light on their underlying cognitive processes. This approach not only aids in making complex theoretical concepts more accessible but also provides educators with a novel method to assess and support student learning in advanced physics education.

This article discusses the significant risks associated with AI-enabled image fraud in scientific publications, particularly focusing on inappropriate duplication and fabrication of images. The paper highlights the widespread implications of such misconduct across various scientific fields and suggests potential preventive measures to combat this growing threat.

  • Joynt, V., Cooper, J., Bhargava, N., & et al. (2024). A comparative analysis of text-to-image generative AI models in scientific contexts: A case study on nuclear power. Scientific Reports, 14, 30377. https://doi.org/10.1038/s41598-024-79705-4

This article evaluates the potential of generative AI to enhance public engagement and literacy concerning low-carbon energy sources, particularly nuclear energy. It finds that while models like DALL-E, DreamStudio, and Craiyon show promise, they are limited by their failure to accurately depict technical details, existing biases, and the representation of indigenous landscapes, highlighting the need for more specialized generative tools.

  • Kupferschmidt, C., Binns, A. D., Kupferschmidt, K. L., & Taylor, G. W. (2024). Stable rivers: A case study in the application of text-to-image generative models for Earth sciences. Earth Surface Processes and Landforms, 49(13), 4213-4232. https://doi.org/10.1002/esp.5961

This article investigates the use of text-to-image generative models in the Earth sciences, specifically evaluating biases in the field of fluvial geomorphology using Stable Diffusion (v1.5). The study highlights that while these models can generate realistic river images with proper prompting, they also exhibit significant training data biase, emphasizing the need for careful review and mitigation in sensitive applications.

  • Rillig, M. C., Mansour, I., Hempel, S., Bi, M., König-Ries, B., & Kasirzadeh, A. (2024). How widespread use of generative AI for images and video can affect the environment and the science of ecology. Ecology Letters, 27, e14397. https://doi.org/10.1111/ele.14397

This article examines the impact of generative AI on ecological research and environmental science, identifying opportunities for improved communication and efficiency as well as risks like bias and misinformation. It emphasizes the need for careful consideration and domain-specific reviews to maximize the benefits while mitigating potential harms.

  • Zhu, L., Lai, Y., Mou, W., & et al. (2024). ChatGPT’s ability to generate realistic experimental images poses a new challenge to academic integrity. Journal of Hematology & Oncology, 17, 27. https://doi.org/10.1186/s13045-024-01543-8

This article discusses the risks posed by AI image generation, particularly concerning the creation of experimental result images in biomedical research. It highlights the need for immediate actions such as restricting experimental image generation, developing detection tools, and adding "invisible watermarks" to maintain academic integrity.