While President Lee Jae-myung and Minister of Economy and Finance nominee Koo Yoon-cheol have been emphasizing the artificial intelligence (AI) sector, the National Statistical Office is also strengthening its focus on AI. The National Statistical Office aims to establish an environment where artificial intelligence can directly read and interpret statistical data, and to transform itself into a "data control tower."
According to the National Statistical Office on the 7th, it has established the autonomous organization "Artificial Intelligence Statistical Innovation Division" as of the 30th of last month and begun operations. The Artificial Intelligence Statistical Innovation Division plans to build metadata for 1,400 statistics by 2028 to enable artificial intelligence to access the National Statistical Office's databases and utilize raw data.
Metadata is "data that describes data." The statistical figures themselves make it difficult to understand "what, how, and when was measured"; however, metadata plays an essential role in creating documentation that enables artificial intelligence to find and interpret data.
Once the metadata is established, it is expected to reduce the likelihood that artificial intelligence will produce inaccurate figures by relying on secondary sources such as news articles, blogs, and reports.
The Artificial Intelligence Statistical Innovation Division plans to promote the establishment of two types of metadata, including economic activity surveys and regional employment surveys, by the end of this year. Once this work is complete, artificial intelligence will be able to accurately answer questions such as "Which region has the highest unemployment rate in our country?", "What differences exist in employment rates by gender?", and "When was the highest increase in unemployment rate by year?"
This is not the first attempt by the National Statistical Office to connect artificial intelligence with statistics. Historically, the National Statistical Office has utilized artificial intelligence in three tracks: automation of statistical classification that was previously done manually, services that allow users to easily access statistics using artificial intelligence, and the production of experimental statistics through artificial intelligence and big data.
The National Statistical Office confirmed during this process that the ability of artificial intelligence to read and interpret statistical data is essential, which led to the promotion of this policy.
An official from the National Statistical Office noted, "Instead of training artificial intelligence with data, it is about enabling artificial intelligence to access the statistical database directly to provide accurate results," and added, "This could reduce incorrect or potentially misleading results (AI hallucinations)."