The role of libraries in the age of artificial intelligence: an analytical look at emerging AI technologies and their applications
Keywords:
artificial intelligence, digital uncertainty, information noise, generative neural networks, intelligent search engine, semantic search, verification, scientific heritage, large language models, Retrieval Augmented GenerationAbstract
This comprehensive study presents a fundamental analysis of the transformation of the functional role of modern scientific libraries in the context of global «digital uncertainty» and the exponential development of generative artificial intelligence technologies. The authors conduct a deep retrospective review of the evolution of neural network architectures – from Frank Rosenblatt’s first probabilistic perceptron models and recurrent networks that solved the problem of long-term memory, to modern Transformers and Large Language Models, justifying the inevitability of the current technological transition. The study systematizes advanced generative AI tools, including diffusion visualization models (Stable Diffusion), speech recognition technologies, and document structure understanding, with a detailed assessment of the prospects for their implementation in cultural heritage preservation processes. Special emphasis is placed on the risks of «information noise», neural network hallucinations, and the blurring of the concept of authorship, which updates the library’s new mission as a guarantor of verified knowledge in accordance with the International Federation of Library Associations principles. The practical significance of the research lies in the detailed technical description of the experience of the Central Scientific Library of the Irkutsk Institute of Chemistry named after A.E. Favorskii of the Siberian Branch of the Russian Academy of Sciences in developing an autonomous intelligent search system. The architecture of the solution based on the Retrieval Augmented Generation methodology, local large language models, and efficient fine-tuning methods is presented, ensuring deep semantic search across chemical collections while maintaining full data sovereignty and answer verifiability.
References
Rosenblatt F. The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain // Psychological Review. 1958. Vol. 65. Iss. 6. P. 386–408.
Николенко С., Кадурин А., Архангельская Е. Глубокое обучение. Погружение в мир нейронных сетей. СПб. : Питер, 2018. 480 с.
Hochreiter S., Schmidhuber J. Long Short-Term Memory // Neural Computation. 1997. Vol. 9. Iss. 8. P. 1735–1780.
LeCun Y., Bengio Y., Hinton G. Deep Learning // Nature. 2015. Vol. 521. Iss. 7553. P. 436–444. DOI 10.1038/nature14539.
Attention Is All You Need / A. Vaswani, N. Shazeer, N. Parmar et al. // 31st Conference on Neural Information Processing Systems (NIPS 2017). Long Beach, 2017. DOI 10.48550/ARXIV.1706.03762.
Transformer – новая архитектура нейросетей для работы с последовательностями // Habr : сайт. URL : https://habr.com/ru/articles/341240/ (дата обращения: 05.08.2025).
Тихомиров М.М. Большие языковые модели // ИСП РАН : сайт. URL : https://tpc.ispras.ru/wp-content/uploads/2023/12/lecture14-2023.pdf (дата обращения: 05.08.2025).
Chain of Thought Prompting Elicits Reasoning in Large Language Models / J. Wei, X. Wang, D. Schuurmans et al. // 36th Conference on Neural Information Processing Systems (NeurIPS 2022). Long Beach, 2022. DOI arxiv.org/pdf/2201.11903v1.
High-Resolution Image Synthesis with Latent Diffusion Models / R. Rombach, A. Blattmann, D. Lorenz et al. // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans, LA, 2022. P. 10674–10685. DOI 10.1109/CVPR52688.2022.01042.
Zhang L., Rao A., Agrawala M. Adding Conditional Control to Text-to-Image Diffusion Models // Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). Paris, 2023. P. 3836–3847. DOI 10.1109/ICCV51070.2023.00355.
LoRA: Low-Rank Adaptation of Large Language Models / E.J. Hu, Y. Shen, P. Wallis et al. // International Conference on Learning Representations (ICLR). 2022. DOI arxiv.org/abs/2106.09685.
Robust Speech Recognition via Large-Scale Weak Supervision / A. Radford, J.W. Kim, T. Xu et al. // Proceedings of the 40th International Conference on Machine Learning (ICML). Honolulu, 2023. URL : https://cdn.openai.com/papers/whisper.pdf (дата обраще-ния: 04.08.2025).
Земсков А.И., Телицына А.Ю. Демонстрация возможностей чата GPT в библиотечной деятельности // Научные и технические библиотеки. 2024. № 4. С. 131–145.
IFLA Statement on Libraries and Artificial Intelligence // IFLA : сайт. URL : https://repository.ifla.org/items/8c05d706-498b-42c2-a93a-3d47f69f7646 (дата обращения: 05.08.2025).
Шрайберг Я.Л., Волкова К.Ю. Вопросы авторского права в отношении произведений, созданных при помощи генеративного искусственного интеллекта // Научные и технические библиотеки. 2025. № 2. С. 115–130.
Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks / P. Lewis, E. Perez, A. Piktus et al. // 34th Conference on neural information processing systems NeurIPS. 2020. URL : https://arxiv.org/pdf/2005.11401v1 (дата обращения: 05.08.2025).
Llama 2: Open Foundation and Fine-Tuned Chat Models / H. Touvron, L. Martin, K. Stone et al. DOI 10.48550/arXiv.2307.09288.
Йылмаз Б. Культура чтения в цифровом мире // Книга. Чтение. Медиасреда. 2024. Т. 2. № 1. С. 17–26.