Hand-drawn portraits of Daniel J. Boorstin by Jason Yi-Bing Lin.

 

By Dr. Jason Yi-Bing Lin

Former Harvard President Derek Bok is often quoted by universities trying to justify the cost of tuition: “If you think education is expensive, try ignorance.” But does that statement really hold up? It's a question worth exploring. Daniel J. Boorstin, former Librarian of Congress, offered a different perspective. He argued that the greatest barrier to discovery is not ignorance — it is the illusion of knowledge. As he put it: “Some of history’s greatest discoverers struggled against what was already ‘known.’ Overcoming this illusion of knowledge requires not only profound understanding but also immense imagination to break free from existing frameworks.”

As AI technology continues to expand across industries, one of the growing concerns is the increasing risk of falling into the trap of “AI hallucinations.” Today’s widely used AI models, whether analytic or generative, rely heavily on accurately labeled data for training. Yet, many educational institutions still overlook the foundational importance of data when teaching AI. As a result, students often lack a clear understanding of the data they use. In some cases, they mislabel it or skip proper validation altogether, rushing to feed flawed inputs into AI models. This leads to the well-known issue of “garbage in, garbage out.” Such careless practices not only waste computing power and energy but also reinforce false confidence in AI-generated results—deepening the problem of AI hallucinations. Over time, this contributes to rising educational costs. The key to unlocking Taiwan’s potential for a breakthrough in AI lies in equipping students with the skills to understand, evaluate, and appropriately utilize accurate data—both qualitative and quantitative.

 

Taiwan possesses significantly less data available for AI training compared to countries like the United States and China. On top of that, inconsistent data formats and limited interoperability pose major challenges to the advancement of AI technologies. In response, the Ministry of Agriculture has made standardization of data formats a key priority in its efforts to advance smart agriculture—enabling seamless communication between various agricultural IoT devices. It’s a forward-looking initiative that shows real foresight. In the private sector, many companies are also placing greater emphasis on data science. For example, Winbond Electronics has incorporated data science into its onboarding program for new employees, helping them understand the importance of effectively managing and utilizing company data. I was honored to be invited as a guest lecturer for one of these sessions.

 

The Small and Medium Enterprise Administration (SMEA) under the Ministry of Economic Affairs has recently taken proactive steps to support traditional industries in adopting AI by launching training programs and inviting experts and scholars to develop the course materials. However, after reviewing the initial curriculum draft, the administration found it too heavily focused on technical computer science concepts, making it difficult for participants without an IT background to engage with. After multiple rounds of discussion, we concluded that the course should focus primarily on helping participants understand data rather than diving into the technical foundations of AI models. Take a textile factory, for example—what employees need is the ability to interpret data generated by machines, annotate it accurately, and identify the right AI tools to address practical, real-world challenges. When introducing these tools, instructors shouldn’t expect workers to learn programming, fine-tune AI models, or adjust complex settings like hyperparameters—doing so would only create unnecessary barriers. Instead, the course should focus on intuitive, user-friendly AI tools that let participants input domain-specific data through simple interfaces and get meaningful results, all without writing a single line of code.

In conclusion, for AI empowerment to take root across diverse non-IT industries, employees must be trained to understand the data relevant to their field and equipped with AI tools that require no programming skills. It’s the responsibility of AI experts to create these no-code solutions—ideally as intuitive and easy to use as ChatGPT. Only when professionals in various industries truly understand their own data can AI be applied effectively, without falling into the illusion of knowledge.