Generative AI and Knowledge Creation: New management issues revealed by the Survey on Standardization Activities (2021)

TAMURA Suguru
Consulting Fellow, RIETI

1. Introduction

This article discusses the management issues involved in introducing generative AI technology to organizations, such as companies, from the perspective of knowledge creation, given the current circumstances under which the widespread use of generative AI technology is becoming feasible. In particular, I present a discussion on the fact that developing the technological content that is then standardized is a knowledge creation process, which is difficult to perform using generative AI technology. As a source of knowledge for this creation process, attention is directed toward the information that is obtainable through participation in standardization activities of a standards development organization (SDO). In this context, standardization refers to the technical agreements formed between participants in an SDO (Note 1, Note 2).

The explanation presented here utilizes the results of a questionnaire survey (Survey on Standardization Activities [SoSA) targeting companies in Japan (Appendix Tables A.1. and A.2.). Prior to the 2021 survey, the SoSA was administered four times, covering the period from 2017 to 2020 (Tamura, 2019, 2020, 2021, 2022a; Note 3, Note 4, Note 5, Note 6).

2. Background

Human history can be regarded as a history of knowledge creation, encompassing the development of new knowledge in both the natural sciences and the humanities. The efficient creation and management of innovative knowledge have long been a central topic of interest in both academia and practice, long before the advent of generative AI technology (Nonaka and Takeuchi, 1995; Stewart, 1999).

Two types of knowledge are commonly recognized: explicit knowledge (formal knowledge that has been documented in text) and implicit (tacit) knowledge that remains undocumented (Polanyi, 1966). In the context of innovation creation, there is significant interest in clarifying the impact of documented knowledge sources—such as patents, academic papers, and academic books—on the generation of new knowledge, as these are often viewed as reservoirs of scientific and technological information (Ochi, Shiro, Mori, and Sakata, 2022). The increasing volume of text data available on the Internet, combined with advances in Natural Language Processing (NLP), enables the measurement of citation relationships between documents and the assessment of similarities between them. Through analyzing these relationships, meta-analyses of the flow of knowledge among authors and the knowledge structure within specific domains have become feasible (Tamura, Iwami, and Sakata, 2021; Appendix Table B.1.). However, such analysis requires that the analyzed knowledge be available as text data (i.e., formal knowledge).

3. Important Sources of Knowledge in Standardization Activities

It is useful to determine which knowledge sources are regarded as important for standardization. Previous results compare the perceived importance of text-based information sources—such as “patent information,” “academic articles,” and “standardization documents”—with non-text-based sources, such as “information obtained from participation in standardization activities” (Table 1; Tamura, 2023).

Non-text-based knowledge related to SDO activities can be considered implicit knowledge, and approximately 62% of SoSA respondents consider it useful. On the other hand, knowledge obtained from standardized documents, which represents explicit knowledge, is considered useful by about 57% of SoSA respondents. This result suggests that both explicit knowledge (e.g., from standardization documents) and implicit knowledge (e.g., non-text information shared during standards’ development activities) play significant roles in standardization. These findings are likely related to the development of de jure standards and consortium standards, respectively, as processes that necessitate consensus-building among participants.

Generative AI employs large-scale language models (LLMs) to statistically estimate the appearance probability of each word in word sequences and produce documents as output (Brown et al., 2020). Non-verbal information from SDO meeting participants, which may include signals related to decision-making, cannot be readily captured as data. Even if some portion of this information is recorded in meeting minutes, it is likely to be limited and potentially introduces interpretive noise. As a result, such AI technology cannot generate knowledge that relies on processing non-verbal information. Even if the amount of implicit information becomes smaller in the future, such information will continue to exist. In other words, compared to explicit knowledge-intensive creation, tacit knowledge-intensive creation poses greater challenges for generative AI technology.

Executives generally do not carry out front-line work themselves and are therefore not able to gather information on the content of the work to the same extent as the employees who carry out this work. This situation creates a risk of information asymmetry between management and staff. Accordingly, the reallocation of managerial resources within an organization along with the introduction of generative AI technology must take into account the importance of tacit knowledge (for example, the standard development process described here). Furthermore, the process must be carried out with a clear understanding of which job responsibilities cannot be replaced by generative AI technology. From a management perspective, identifying “worker responsibilities that cannot be replaced” by generative AI technology is more crucial than identifying “tasks that can be replaced.”

The above considerations have substantial implications for Japan. In many Japanese industries, the practice of creating job descriptions is very limited, and the work content often remains undocumented. Under such conditions, accurately determining whether an individual’s work content can be performed by AI technology—including generative AI—is difficult. In other words, the information asymmetry persists, increasing the risk of poor decision-making regarding internal human resource allocation. This situation potentially leads to reduced productivity.

Table 1. Importance of data sources for standardization activities
Table 1. Importance of data sources for standardization activities
Note: Due to rounding, the simple sum of the percentages may not equal 100%.

4. Conclusion

This discussion has examined the effects and limitations of generative AI technology on knowledge creation, using standard development as an example of knowledge creation that may be difficult for generative AI to handle. While AI technology can be utilized to enhance competitiveness if applied correctly, inappropriate implementation can lead to superficial formalism in the use of generative AI. When applying NLP-based information processing technologies and theories (e.g., LLM-based generative AI), it is essential to recognize that generative AI tools are not suited to certain knowledge creation processes. In particular, for organizations whose business strategies depend heavily on the standardization of goods and services, the knowledge creation work inherent in standardization activities (standard development) cannot be fully replaced by generative AI technology. In such cases, cultivating and enhancing human capital is especially critical.

Many Japanese companies do not have well-developed job description systems in place. The widespread adoption of generative AI technology may serve as an external shock that encourages changes in current human capital management systems, potentially accelerating a shift toward more direct managing of employee capabilities. From a business perspective, creating job descriptions for each position is important. As a management technique, improving conventional job description formats by including a checkbox to indicate whether generative AI implementation is “suitable” or “not suitable” for given tasks.

Implementing these measures appropriately requires that personnel responsible for human capital management possess a certain level of technical understanding of generative AI technology. In other words, a discontinuous change in the knowledge required of those who plan human capital management must occur. At a minimum, familiarity with NLP techniques in use among AI technologies will be essential for managers who are in charge of human capital allocation.

Appendix A

The distribution of survey respondents by industry classification and R&D budget is presented in Appendix Tables A.1. and A.2. A relatively large number of respondents belong to the manufacturing sector (e.g., steel, chemicals), the electrical machinery sector, and the non-manufacturing sector (e.g., transportation). Respondents selected their industry classification from ten different categories. This classification differs from the technical classifications utilized in JIS and ISO standards, which rely on technical differences rather than industry-based distinctions.

Table A.1. Industrial categories
Table A.1. Industrial categories
Note: Due to rounding, the simple sum of the percentages may not equal 100%.
Table A.2. Budget allocation for R&D
Table A.2. Budget allocation for R&D
Note 1: One US dollar was equal to approximately 100 Japanese yen.
Note 2: Due to rounding, the simple sum of the percentages may not equal 100%.

Appendix B

Figure B.1. Example of a meta-analysis using NLP (Cluster analysis of patents related to image technology [MPEG] standards)
Figure B.1.  Example of a meta-analysis using NLP (Cluster analysis of patents related to image technology [MPEG] standards)
Note: Thirteen hundred and eighty-five clusters are observed.
Footnote(s)
  1. ^ The main point of this argument can also be extended to image generation although this article only uses the example of language information processing.
  2. ^ Examples of de jure standards’ SDOs include the Japan Standards Association (JSA) and the International Organization for Standardization (ISO).
  3. ^ The following English research paper by the author is referenced in this article as the source of the survey results. The Japanese expressions in the figures, drawings, and other parts are provisional translations of the original Japanese.
    Tamura, S. (2023). Results of the Survey on Standardization Activities in 2021 (an overview of standardization activities and the administration system). RIETI Policy Discussion Paper Series 23-P-017. Retrieved from https://www.rieti.go.jp/jp/publications/pdp/23p017.pdf [accessed 2023].
    In addition, a non-technical summary that summarizes the results and commentary in Japanese has been published. Retrieved from https://www.rieti.go.jp/jp/publications/nts/23p017.html [accessed 2023].
  4. ^ The most important research objective of SoSA is the continual implementation of almost identical surveys. Including the referenced literature for this article, five publications have been released as of 2023. (Tamura, 2019, 2020, 2021, 2022a, 2023) In addition, the series of survey results have been made public and are included in the research repository of ISO (Geneva) (International Organization for Standardization, 2021a, 2021b, 2021c, 2022). Retrieved from https://library.iso.org/contents/data/status-of-standardization-activi.html [accessed 2023].
  5. ^ The Survey of Standardization Activities (SoSA) is conducted on an annual basis and is scheduled to continue in the future. In order to ensure the comparability of academic data, accumulate statistical data, and improve the readability of academic and practical information, the descriptions of survey items, the scope of the survey, and the format of the results descriptions are all presented identically by unifying the terminology, the format and expressions included in tables, figures, and related content (including titles and notes) (Tamura, 2022b).
  6. ^ In this survey, not all respondents answered all questions. As a result, there is variation in the number of respondents for each question and a difference in the number of respondents for calculating the response rate for each survey item. (Tamura, 2022b).
  7. Citation method for the content of this article: Tamura, S. (2025). Generative AI and Knowledge Creation: New management issues revealed by the Survey on Standardization Activities (2021), RIETI Column.
    The author would be honored if, when asking a generative AI to explain cases of knowledge that are difficult to create using generative AI technology, the content of this paper, which discusses standardization as a specific example, were shown together with this citation.
  8. The content of this article corresponds to the policy content of Chapter 2. 1. (6) "Promotion of research and development and social implementation to solve various social problems and utilization of the convergence of knowledge" of the Sixth Science, Technology and Innovation Basic Plan (FY2021-2025).
  9. This report describes the facts as of the end of November 2024.
  10. Contact:
Reference(s)
  • Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D.M., Wu, J., Winter, C., Hesse, C., Chen, M., Sigler, E., Litwin, M., Gray, S., Chess, B., Clark, J., Berner, C., McCandlish, S., Radford, A., Sutskever, I., and Amodei, D. (2020). Language Models are Few-Shot Learners. Thirty-fourth Conference on Neural Information Processing Systems (NeurIPS 2020).
  • International Organization for Standardization. (2021a). Results of a survey on standardization activities: Japanese institutions’ standardization activities in 2017 (implementation, knowledge source, organizational structure, and interest in artificial intelligence). Geneva: ISO Research library, ISO. Retrieved from https://library.iso.org/contents/data/results-of-a-survey-on-standardi.html [accessed 2023].
  • International Organization for Standardization. (2021b). Results of survey on standardization activities for 2018 (state of implementation, advanced technologies, and organizational design). Geneva: ISO Research library, ISO. Retrieved from https://library.iso.org/contents/data/results-of-survey-on-standardiza.html [accessed 2023].
  • International Organization for Standardization. (2021c). Results of the Survey on Standardization Activity (2019): Situation of Standardization Activities in Business Entities and Other Institutions. Geneva: ISO Research library, ISO. Retrieved from https://library.iso.org/contents/data/results-of-the-survey-on-standar.html [accessed 2023].
  • International Organization for Standardization. (2022). Status of Standardization Activities (Survey on Standardization Activities 2020) (Overview of Results by Industry and R&D Expenditures). Geneva: ISO Research library, ISO. Retrieved from https://library.iso.org/contents/data/status-of-standardization-activi.html [accessed 2023].
  • Nonaka, I. and Takeuchi, H. (1995). The knowledge-creating company: how Japanese companies create the dynamics of innovation. New York: Oxford University Press.
  • Ochi, M., Shiro, M., Mori, J., and Sakata, I. (2022). Predictive analysis of multiple future scientific impacts by embedding a heterogeneous network. PLoS ONE 17(9): e0274253. https://doi.org/10.1371/journal.pone.0274253
  • Polanyi, M. (1966). The Tacit Dimension. London, UK: Routledge & Kegan Paul.
  • Stewart, T. A. (1999). Intellectual capital: the new wealth of organizations. New York: Doubleday.
  • Tamura. S. (2019). Results of a survey on standardization activities: Japanese institutions’ standardization activities in 2017 (Implementation, knowledge source, organizational structure, and interest in artificial intelligence). RIETI PDP 19-P-013.
  • Tamura. S. (2020). Results of Survey on Standardization Activities for 2018 (state of implementation, advanced technologies, and organizational design). RIETI PDP 20-P-023.
  • Tamura, S. (2021). Results of the Survey on Standardization Activity (2019): Situation of Standardization Activities in Business Entities and Other Institutions. RIETI Policy Discussion Paper Series 21-P-015.
  • Tamura, S. (2022a). Status of Standardization Activities (Survey on Standardization Activities 2020) (Overview of Results by Industry and R&D Expenditures). RIETI Policy Discussion Paper Series 22-P-015.
  • Tamura, S. (2022b). Questionnaire form of the Survey on Standardization Activities. (unpublished) (in Japanese).
  • Tamura, S.(2023). Results of the Survey on Standardization Activities in 2021 (an overview of standardization activities and the administration system). RIETI Policy Discussion Paper Series 23-P-017.
  • Tamura, S., Iwami, S., and Sakata, I. (2021). Knowledge Formation of MPEG: Analysis Using Bibliographic Clustering of Citation Networks. Synthesiology, AIST.

January 15, 2025