Do large language models in Taiwan have more of a "Taiwanese flavor"? Making AI understand you better also makes it more trustworthy. Highlights

Generative AI models are competing on all fronts, and the National Science and Technology Council is also promoting a Taiwan version of the AI language model program, allowing rapidly developing AI technology to bring tremendous innovation and benefits. However, the large amounts of AI data also bring new social and humanistic problems and risks. This gave rise to the Taiwan Digital Trust Association. So, where is AI development heading, and how can we create an era of trustworthy AI?

Host: Ethan Liu Jiezhong

Guests:

— Hsu Yongzhen, Deputy Chief Convener, Taiwan AI Excellence Center

— Chan Tingyi, Chairperson, Artificial Intelligence Technology Foundation

Key Highlights

"Will Taiwan's Large Language Models Have More 'Taiwanese Flavor'? Making AI Understand You Better and Be More Trustworthy"

Advantages and Strategies for AI Development in Taiwan

Chan Tingyi, chairperson of the Artificial Intelligence Technology Foundation, pointed out that Taiwan's semiconductor industry holds a core position in the global supply chain, providing an indispensable hardware foundation for AI development. Additionally, Taiwan's years of accumulation in manufacturing technology have laid a solid foundation for AI applications. She emphasized that AI development is not just about technological progress; the key lies in value creation through application scenarios. Especially in Taiwan's digital transformation process, AI acts as an enabling engine, helping various industries redesign processes and optimize services.

Hsu Yongzhen, deputy chief convener of the Taiwan AI Excellence Center, mentioned that since 2023, Taiwan has actively invested in the development of local language models. This initiative is not only driven by technological advancement considerations but also by the need for data sovereignty and the preservation of linguistic culture. He stressed that Taiwan faces similar AI development challenges; as an example, ChatGPT's success has raised public awareness of AI but has also highlighted Taiwan's lack of AI technological autonomy. If Taiwan overly relies on foreign technology, it may be constrained by external forces, making it difficult to control the direction of key technological developments. Taiwan's language model projects not only help enhance digital competitiveness but also reduce the risks associated with Simplified Chinese information bias.

Taiwan's continued investment in AI research, including adopting an open architecture strategy despite limited national resources, aims to absorb foreign technologies while meeting domestic industry needs. AI has broad application prospects in healthcare, manufacturing, and other fields, with unique advantages in solving localization problems. AI development is like a marathon, requiring systematic planning of technological advancement steps to ensure long-term international competitiveness.

The importance of building digital trust

Digital trust is a core element of AI technology applications. As AI technology penetrates into every corner of life, whether it is governments, businesses, or consumers, data governance is needed to ensure system transparency and fairness. AI developers and service providers should be responsible, ensure the transparency and accountability of AI systems, and try to reduce the risks posed by technology.

Zhan Tingyi, chairman of the Artificial Intelligence Technology Foundation, believes that digital trust is the cornerstone of the digital economy and network governance. He emphasized that companies should take responsibility when providing AI products and services to ensure that the development and application of AI comply with social ethics and values. Regulate the development and application of AI, and hierarchically manage the risks of AI products and services to ensure their security and trustworthiness.

Hsu Yong-Chen, vice president and convener of the Taiwan AI Center of Excellence, also agreed with the importance of digital trust. He believes that in the AI era, individuals cannot completely rely on the state or platform to protect their own rights and interests, but should have the ability to make independent judgments and choices. He suggested using technical means, such as giving true and false information different speeds of dissemination, to help people increase their response to bad information.

Talent Development in the AI Era  

Both speakers emphasized that the cultivation of AI talent should extend to cross-disciplinary knowledge, as in the process of digital transformation, the connections between processes are extremely important. When processes are linked, they create assistance. Whether in government, enterprises, or organizations, it is particularly important for talent to cover knowledge from various industries. With the application of digital transformation across industries, the development and application of AI technology rely heavily on the support of talent. AI not only requires technical experts but also experts familiar with different industry domains to help realize the value creation of digital transformation.  

Chairperson Jane Chan of the Artificial Intelligence Technology Foundation pointed out that, besides manufacturing being crucial for creating GDP, AI should be well utilized in digital transformation across different fields to generate higher value. It is also important to consider how AI can help cultivate experts in various domains; Taiwan needs to develop cross-disciplinary AI talent. Future AI talent not only needs to be proficient in programming and algorithms but also must possess cross-disciplinary knowledge and skills to have a holistic view. This mindset in implementation will ensure that our use of AI can be more comprehensive, applying AI technology across all fields.  

Vice Convenor Ernest Hsu of the Taiwan AI Excellence Center approached the topic from an educational perspective, noting that AI is a very convenient tool. It can help you find, organize, and analyze information as well as answer questions, but final judgment still rests with humans. Therefore, the user's understanding and judgment of information are especially crucial. He also emphasized the importance of AI literacy, recommending the strengthening of 'critical reading' education in the AI era to analyze information sources and cultivate independent thinking. This ability is not only important when dealing with AI-generated information but also helps prevent being misled by inappropriate information. Ernest Hsu also believes that talent development should focus on enabling people to remain calm amidst the flood of information and acquire skills to handle complex situations, allowing them to better adapt to life and work in the AI era.

Media Literacy and Information Verification Skills

With the rapid proliferation of generative AI, Wen-Chen Hsu, Deputy Convener of the Taiwan AI Excellence Center, emphasized that media literacy should be part of education. Although generative AI possesses fluent language capabilities, the content it produces can contain serious errors or even fabricated information. Therefore, the public needs to learn the skills to discern true from false information and should not overly rely on algorithmic recommendations to avoid being misled by false information.

Ting-Yi Chan, Chairperson of the Artificial Intelligence Technology Foundation, suggested that when launching services, products, or information, people should also understand the data sources behind them, using technical tools to help users improve their ability to verify information as a basis for judgment. The two speakers also mentioned that media literacy and information literacy courses are increasingly being offered in both education and the corporate sector. This kind of "preventive education" can help the public better navigate the complexity of the information environment, allowing them to selectively process information they receive and absorb, and avoid relying solely on a single source or platform.

AI Governance and Legislation

Chairperson Zhan Tingyi of the Artificial Intelligence Technology Foundation mentioned that when Taiwan promotes the “Basic Law on Artificial Intelligence,” it can take inspiration from the European Union's tiered management system, handling AI through a risk-based classification approach. Regarding how large enterprises should be regulated, traditional legislative models are no longer suitable for the internet era. Regulations need to align with the inherent nature of the internet to establish highly practical and enforceable rules. For example, high-risk applications in areas such as healthcare and finance should comply with stricter regulations to protect users' rights. She believes that risk-based management of AI applications helps clarify the allocation of responsibilities for different types of AI technologies and their application areas, thereby ensuring public safety.

AI governance should advance in parallel with digital governance to holistically enhance Taiwan’s adaptability in the digital transformation era. She emphasized that technological development is not only a matter of the technology itself but also involves collaboration among various stakeholders. From enterprises to the government, the healthy development of AI technology requires joint efforts to ensure that the direction of AI development aligns with the interests of society as a whole.

Additionally, Deputy Convener Xu Yongzhen of the Taiwan AI Excellence Center pointed out that drafting the AI Basic Law is only the first step. It is necessary to establish a comprehensive legal framework for technology development and data governance. Taiwan should consider both the characteristics of AI technology development and its potential risks, avoiding the limitations of a single governance approach. While the AI Basic Law provides the foundation for promoting safe and responsible AI use, it is even more critical to create a regulatory framework adaptable to technological changes. This framework should include key elements such as information transparency, accountability, and data privacy protection.

Both speakers agree that as Taiwan develops AI, it must take a multi-faceted approach: developing autonomous technology, establishing a trustworthy environment, cultivating AI talent, enhancing media literacy, and formulating comprehensive AI governance regulations. Only then can Taiwan secure a strong position in the AI era.