Recently, many major countries around the world, starting with the U.S., Japan, Germany, China, U.K., etc., have issued an administrative order to ensure the safety of AI technology, putting an emphasis on the safe, effective implementation of AI into their systems. In line with such trends, Korean researchers have collaborated with renowned AI experts from all around the world to create new AI-related international standards, garnering attention from the global AI community.
Proposal No. |
Title |
Status |
ISO/IEC 5259-1 |
Artificial intelligence - Data quality for analytics and ML - Part 1: Overview, terminology, and examples |
Enacted |
ISO/IEC 5259-2 |
Artificial intelligence - Data quality for analytics and ML - Part 2: Data quality measures |
In Progress |
ISO/IEC 5259-3 |
Artificial intelligence - Data quality for analytics and ML - Part 3: - Data Quality Management Requirements and Guidelines |
Enacted |
ISO/IEC 5259-4 |
Artificial intelligence - Data quality for analytics and ML - Part 4: Data quality process framework |
Enacted |
ISO/IEC 5259-5 |
Artificial intelligence - Data quality for analytics and ML - Part 5: Data quality for analytics and machine learning (ML) - Part 5: Data quality governance |
In Progress |
ISO/IEC TR 5259-6 |
Artificial intelligence - Data quality for analytics and ML - Part 6: Visualization framework for data quality |
In Progress |
Korean researchers have successfully created an international standard that can effectively evaluate and manage the quality of data used in the AI development process to support the stability and reliability of emerging AI-related technologies and services.
Therefore, as the quality of data used in AI development can be measured and evaluated in the future through a set of standardized standards, the way to easily identify the quality of a specific set of data has been opened. Now, not only customers can determine whether to purchase the service or not based on its quality, but developers and providers of AI-based services can also use these standards to constantly improve the quality of the data used in the AI development process to further enhance their services.
Electronics and Telecommunications Research Institute (ETRI) announced that the 『Data Quality Standard for Data Analysis & Machine Learning』, which has been developed to promote the international standardization in the AI, has finally been established as an official international standard. It is known that the leader of this project is Ha Suwook, the Technical Director of ETRI's Strategic Standards Research Section.
The research team established an overview and common concept regarding the 'quality of data' used for the development and machine learning of AI in the 『Data Quality Standard for Data Analysis & Machine Learning』, a series consisting of six different sections. In particular, the team has led the development of "Part 1: Overview, Terms, Examples," to help understand the basic concepts and implementation of the series/standards. Part 1 includes the following concepts: ▲Measurement of Data Quality ▲Requirements for Data Quality Management ▲Procedures for Data Quality Management, etc.
These standards provide a set of reliable tools and methods for organizations/institutions to evaluate, manage, and improve the quality of data use in the process of AI development, which is crucial in the modern era of data-based decision-making. It will help these organizations effectively use data based on their varying purposes, while providing the definition of common terms used in the field along with reliable action plans that suit the current industry environment.
The standards established by the research team of ETRI focuses on the following aspects: ▲Ensure data quality to enable the development of safe AI models while minimizing errors and biases ▲Increase the reliability of AI systems to maintain the consistency of its performance while supporting effective management of data lifecycle ▲Strengthen AI safety to help risk management and support legal compliance.
In particular, this can be considered a significant feat as it is the world's first official standard to support AI safety and reliability from a data quality perspective.
Through this, the new standard is expected to help secure a high level of reliability and stability in the field of AI, as well as reducing the frequency of legal disputes and easing monitoring activities conducted by regulatory agencies. In addition, it is also expected to increase the interoperability between companies and promote the development of innovative AI application services based on high-quality data.
Based on this achievement, the research team will put their eyes on the development of another international standard regarding data requirements, utilized information, shared information for machine learning models, etc. according to the data lifecycle in an AI development environment to support AI reliability and stability. To this end, ETRI is planning to continue its cooperation with experts from around the world (U.S., Japan, Germany, China, U.K., etc.) as well as domestic industry/academic experts, with a focus on JTC 1 SC42 WG 2.
Lee Seung-yun, the head of ETRI's Standards & Open Source Research Division, said, "This standard has been officially established and accepted as an international standard* thanks to the active support from member states. It will provide an important criterion for evaluating the quality of data used for AI development, and is expected to make meaningful contributions in improving the overall reliability and performance of AI technology. It is also a groundbreaking feat in a domestic level considering that South Korea is working on the establishment of its own AI Safety Research Institute."
* Refers to ISO/IEC 5259-1
The research team predicts that this new international standard will play a pivotal role in laying the foundation for responsible development and distribution of AI systems and further strengthening the cooperation among various stakeholders.
