The Transformation of Digital Scholarship Paradigm in AI2.0 Era

Abstract

Abstract:

Digital scholarship is a research paradigm in which digital technologies and methodsintervene in the academic research process. The breakthrough of generative AI represented by ChatGPTmarks the arrival of generalized AI, which some scholars call the “AI2.0 era”, and also makes scientificresearch transition from data-intensive to computationally-intensive, thus entering the fifth paradigm.More and more of the research process will be involved in, or even dominated by AI. Different from datadrivenresearch, big models not only can conduct data acquisition, extraction, management and analysisin a more efficient, or even “automatic” way, but also have the potential to realize the “understanding” ofmassive scientific data, so as to discover new laws and trends. The big model can not only help scientistsbetter understand the data, but also provide a more comprehensive and in-depth perspective for scientificresearch. At the same time, it can also be applied to a variety of related scenarios, to help scientistsfind errors and defects in the data or results, thereby improving the reliability and repeatability ofscientific research. Of course, with the wide application of AI technology, it also brings some challenges.For libraries, research-oriented services will become more academic and technical. In addition toinfrastructure, construction of research communication environment and data support services, they willintervene more deeply in the research process with AI thinking and digital methods and tools, provide thenecessary support in protecting the privacy and security of data, and avoid data leakage, misuse or ethicalissues.

Jin Jiaqin, Liu Wei (Shanghai Library). The Transformation of Digital Scholarship Paradigm in AI2.0 Era[J]. Libraly Journal, 2023, 42(391): 10-21.

References

[ 1 ] Hey T， Tansley S， Tolle K，等 . 第四范式：数据密集型科学发现[M]. 北京：科学出版社， 2012.

[ 2 ] 邓仲华，李志芳 . 科学研究范式的演化——大数据时代的科学研究第四范式[J]. 情报资料工作，2013（4）：19–23.

[ 3 ] J Shaw. Why “Big Data” Is a Big Deal：InformationScience Promises to Change the World[J]. HarvardMagazine，2014，2（3/4）： 30–35；74–75.

[ 4 ] Jinghai Li， Wenlai Huang.Paradigm shift inscience with tackling global challenges[J]. NationalScience Review， 2019， 6（6）： 1091–1093.

[ 5 ] C Leng，Z Tang，Y-G Zhou，et al. Fifth Paradigmin Science： A Case Study of an Intelligence-DrivenMaterial Design[J/OL]. Engineering， https://doi.org/10.1016/j.eng.2022.06.027.

[ 6 ] Wang H， Fu T， Du Y， et al. Scientific discoveryin the age of artificial intelligence[J]. Nature，2023， 620（7972）： 47–60.

[ 7 ] Wang H， Fu T， Du Y， et al. Artificial intelligencein science： An emerging general method ofinvention[J]. Research Policy， 2022， 620（7972）：47–60.

[ 8 ] Christopher Bishop.AI4Science to empower thefifth paradigm of scientific discovery[EB/OL].[2023-07-10]. https://www.microsoft.com/en-us/research/blog/ai4science-to-empower-the-fifthparadigm-of-scientific-discovery/.

[ 9 ] 张伟 . AI 赋能科研“第五范式”引变革[N]. 中_国高新技术产业导报，2022-08-15（13） .

[10] 王飞跃，缪青海 . 人工智能驱动的科学研究新范式：从AI4S 到智能科学[J]. 中_国科学院院刊，2023，38（4）：536–540.

[11] 王飞跃，缪青海，张军平，等 . 探讨AI for Science的影响与意义：现状与展望[J]. 智能科学与技术学报，2023，5（1）：1–6.

[12] 梁娜，曾燕 . 推进数据密集科学发现提升科技创新能力：新模式、新方法、新挑战——《第四范式：数据密集型科学发现》译著出版[J]. 中_国科学院院刊，2013，28（1）：115–121.

[13] Zhao W X，Zhou K，Li J，et al. A survey of largelanguage models[EB/OL]. [2023-07-17]. arXivpreprint arXiv：2303.18223， 2023.

[14] Terry S. 深度学习：智能时代的核心驱动力量[M].北京：中_信出版社，2019.

[15] Brown T， Mann B， Ryder N， et al. Languagemodels are few-shot learners[J]. Advances inneural information processing systems， 2020， 33：1877–1901.

[16] Radford A， Wu J， Child R， et al. Languagemodels are unsupervised multitask learners[J].OpenAI blog， 2019， 1（8）：9.

[17] 王静静，叶鹰，王婉茹 . GPT 类技术应用开启智能信息处理之颠覆性变革[J]. 图书馆杂志，2023（5）：9–13.

[18] Radford A， Narasimhan K， Salimans T， et al.Improving language understanding by generativepre-training[EB/OL]. [2023-07-17]. https://s3-us-west-2.amazonaws.com/openai-assets/research-covers/language-unsupervised/language_understanding_paper.pdf.

[19] Rober W McGee.Annie Chan：Three Short StoriesWritten with Chat GPT[EB/OL]. [2023-07-17].DOI：10.13140/RG.2.2.21169.66401（2023） .

[20] Zhou J，Ke P，Qiu X，et al.ChatGPT： potential，prospects， and limitations[J]. Front InformTechnol Electron Eng（2023） . DOI：https://doi.org/10.1631/FITEE.2300089.

[21] Wang Y， Yao Q， Kwok J T， et al. Generalizing froma few examples： A survey on few-shot learning[J].ACM computing surveys，2020， 53（3）：1–34.

[22] How Chat GPT Will Revolutionize The Futureof Bakery[EB/OL]. [2023-07-10]. https://www.puratos.com/blog/how-chat-gpt-will-revolutionizethe-future-of-bakery.

[23] Wei J， Tay Y， Bommasani R， et al. Emergentabilities of large language models[J]. Transactionson Machine Learning Research，2022. DOI：https://doi.org/10.48550/arXiv.2206.07682.

[24] Wei J，Wang X，Schuurmans D， et al. Chainof-thought prompting elicits reasoning inlarge language models[J]. Advances in NeuralInformation Processing Systems， 2022， 35：24824–24837.

[25] Silver D，Huang A， Maddison C J，et al. Masteringthe game of Go with deep neural networks and treesearch[J]. nature， 2016， 529（7587）： 484–489.

[26] GPT-4[EB/OL]. [2023-07-10]. https://openai.com/research/gpt-4.

[27] Scott Clark. Unlock the potential of your data with AI：How to break down data silos and extract valuableinsights with AI and NLP technology[EB/OL].[2023-07-10]. https://www.cmswire.com/customerexperience/how-artificial-intelligence-can-breakthrough-data-silos/.

[28] Antoine Tardif. Is GPT-4 a Leap Forward TowardsReaching AGI?[EB/OL]. [2023-07-10]. https://www.unite.ai/is-gpt-4-a-leap-forward-towardsreaching-agi/.

[29] Bubeck S， Chandrasekaran V， Eldan R， et al.Sparks of artificial general intelligence： Earlyexperiments with gpt-4[EB/OL]. [2023-07-10].arXiv preprint arXiv：2303.12712， 2023.

[30] Sergio De Simone. Running Large LanguageModels Natively on Mobile and Laptops[EB/OL].[2023-07-10]. https://www.infoq.com/news/2023/ 05/mlc-llm-mobile-laptops/.

[31] Liz O’Sullivan，John Dickerson. Here are a fewways GPT-3 can go wrong[EB/OL]. [2023-07-10].https://techcrunch.com/2020/08/07/here-are-a-fewways-gpt-3-can-go-wrong/.

[32] 郑一波，曾建勋 . 图书馆发现系统的实质与应对策略[J]. 情报科学，2022，40（8）：134–137；143.

[33] Perplexity[EB/OL]. [2023-07-10]. https://www.perplexity.ai/.

[34] Tang R， Han X， Jiang X， et al. Does syntheticdata generation of llms help clinical textmining?[EB/OL]. [2023-07-10]. arXiv preprintarXiv：2303.04360， 2023.

[35] Jamall Brown. Building a Search Chatbot withGPT-3[EB/OL]. [2023-07-10]. https://medium.com/@jamall.j.brown/enhancing-knowledgeretrieval-with-gpt-3-part-1-c0f0c3b27d01.

[36] 浙商证券 . GPT-4 发布多模态处理+ 复杂推理能力有望打开应用空间[EB/OL]. [2023-07-10]. https://pdf.dfcfw.com/pdf/H3_AP202303161584308890_1.pdf?1678965405000.pdf.

[37] 托马斯·库恩 . 科学革命的结构[M]. 北京：北京大学出版社， 2022.