Taiwan’s AI training data triples under ministry’s efforts - Taipei Times

Taiwan’s Sovereign AI Training Corpus has more than tripled in size since its launch in December last year, with the number of datasets rising to more than 5,000 and the number of tokens reaching 2.1 billion, the Ministry of Digital Affairs said today.

The ministry established the platform last year to compile artificial intelligence (AI) training data that reflects Taiwanese cultural characteristics, Minister of Digital Affairs Lin Yi-ching (林宜敬) said.

The ministry has made government-owned copyrighted materials as widely available as possible and incorporated them into the corpus for use by AI development teams in Taiwan, Japan and the US, Lin said.

Photo: Chiu Chiao-chen, Taipei Times

The corpus covers diverse topics such as Taiwanese folk customs, local languages, ecology, tourism, publications, law, art, cultures, economics, geography, transportation, healthcare and history.

The ministry said it hopes to continue expanding the corpus to provide generative AI developers with the data necessary to develop AI applications that better reflect Taiwanese culture.

As the government’s data collection grows, the next step would be to gather more high-quality original content from the private sector, Department of Data Innovation Director Wang Fu-chung (王復中) said.

The ministry is collecting several types of content, including books and other copyrighted works, with contributors’ consent and the choice to opt out, Wang said.

The data can be used by government agencies, private organizations and academic institutions for model training, fine-tuning, retrieval-augmented generation and knowledge extraction, he said.

It also has applications in education, administrative support, smart customer services and professional consulting, he added.

Publishers are responsible for selecting and checking the quality of privately contributed content, Wang said.

The corpus checks the content structure and data format and considers whether the data can be used for AI training, without interfering with the content itself, he said.

Lin, who is also a writer, was among the first to license his works for use as AI training data, encouraging more creators to take part.

While ensuring creators’ rights, the ministry hopes to continue collecting AI training data that reflects Taiwanese culture, so that the country does not get left out as AI models rapidly develop worldwide, Lin said.

Taiwan’s growing defense preparations might be sending a stronger signal to the US than to Beijing, highlighting the need for Taipei to tailor its deterrence efforts to different audiences, a US security academic said yesterday. Stanford University Freeman Spogli Institute for International Studies fellow Oriana Skylar Mastro made the comment during a Taipei Security Dialogue panel examining lessons Taiwan could draw from the wars in Ukraine and the Middle East. Mastro said she was impressed by Taiwan’s increased defense spending, emphasis on asymmetric capabilities and efforts to prepare society for conflict, but questioned whom those initiatives were influencing. She said when she asked

A total of 400 foreign “high- level” professionals have been naturalized as Republic of China (Taiwan) citizens over the past decade as the government has eased nationality rules to attract talent from a wider range of fields, the Ministry of the Interior said yesterday. The 400th high-level professional, who works in the electronics industry, was approved for ROC citizenship on Aug. 26, the MOI said. The MOI quoted the newly naturalized citizen as saying that Taiwan is a leader in fields such as semiconductors, electronics manufacturing, artificial intelligence development and energy technology. He also said

China has warned that it could cancel a planned US-China leaders’ summit if Washington approves new arms sales to Taiwan, a Japanese media outlet quoted sources as saying yesterday. US President Donald Trump, “apparently eager to showcase some diplomatic achievements ahead of November’s midterm congressional elections, places great importance on holding the summit with Chinese President Xi Jinping (習近平),” Japan’s Kyodo News reported. “While preparing for Xi’s upcoming trip, China has restated its position that the Taiwan issue is a ‘red line’ in bilateral relations,” the report said. The report said that when Xi met with Trump in Beijing in May, Xi said

RARE PEEK: The footage released showed one of the bears staying in a cave for as long as 70 days, providing the first record of Formosan black bear’s overwintering The Forestry and Nature Conservation Agency yesterday released tracking records of released Formosan black bears for the first time, showing that they get much closer to human settlements than previously thought. The agency said it had captured two bears, Da Qu-ali (達古阿里) and Aziman (阿里曼), in indigenous villages in Hualien County and began to track them with a camera-mounted satellite collar after releasing them back into the wild. The results confirmed that the bears have moved from deep mountain areas into lowland hills closer to human communities, leading to an increase in human-bear conflicts, it said. The agency has introduced artificial intelligence-assisted early