A summary of a lecture delivered by a former Google executive and a pioneer in language model research, addressing the profound impact of AI on society, education, and Taiwan’s future.
It’s exciting to be back on campus. Today, I want to talk about a topic that is changing almost weekly: the impact of Artificial Intelligence on your generation. But I want to start from a different angle, one that is critical to our future here. The biggest challenge facing Taiwan today isn’t AI; it’s our declining birthrate.
We are standing at a crossroads similar to Japan 30 years ago. Japan’s economic stagnation over the past three decades was driven significantly by its aging population. But we have a tool Japan didn’t have then: AI. This technology could be the key to ensuring that our next ten years, which demographers predict will be a “golden decade” before the full effects of our aging society take hold, can be sustained long into the future.
Welcome to the AI Generation: Immigrants in a New World
The United Nations has declared that every one of us, from a one-year-old to a ninety-year-old, is now part of the AI Generation. But unlike previous technological shifts, we are not natives here. We are all immigrants in this new era, learning how to coexist with and effectively utilize this powerful technology.
Unlike the mobile internet era, where a toddler could intuitively use a touchscreen, mastering AI requires deep domain knowledge. The more expertise you have in your own field, the better you can judge the quality of the information AI provides. This is a fundamental shift. You are the first generation of beneficiaries in this AI era, and your time on campus is the perfect opportunity to learn how to master this relationship.
The most profound change AI brings is its ability to “speak human language” (說人話). This has led to a paradigm shift from searching to conversing. In the age of Google Search, we typed keywords and read through multiple sources. In the AI era, we ask a question and often accept the first answer as the definitive one. This could lead to a decline in our reading and critical thinking habits, as we outsource the act of synthesis to the machine.
This conversational ability will soon break down language barriers. Within a couple of years, Bluetooth earbuds will offer real-time translation, allowing you to travel the world speaking only your native tongue. The world will be flat in a way we’ve never imagined. The barrier will no longer be language, but culture and mindset. For us in Taiwan, an island nation, we must consciously break out of our “echo chamber” (同溫層) and engage with the world, a task that will become both easier and more critical than ever.
The Power and Peril of Generative AI
Today’s AI can be broadly divided into two categories: Predictive AI, which analyzes data to make a prediction (like facial recognition), and Generative AI, which creates new content based on a prompt. You are most familiar with the latter through tools like ChatGPT.
The challenge of Generative AI is that everyone, from a novice user to the world’s top developer, uses the same interface: natural language. This makes it deceptively simple. You can always get an answer, which makes you think you’ve asked the right question. The real skill lies in understanding the model’s boundaries and learning how to craft prompts that unlock its full potential.
Its capabilities are already staggering. I was one of the first people in Taiwan to research Language Models for my PhD at National Taiwan University. What we did then was rudimentary. Today’s models have ingested the equivalent of 100 university libraries. Their ability to read, summarize, and even reason through “Chain of Thought” processes is rapidly approaching expert human levels. Tools like Deep Research can generate reports on complex topics, complete with citations and logical arguments, that surpass what most humans can produce.
But this power comes with a significant risk. An estimated 98% of the internet’s content could be AI-generated within five years. The current ad-supported model that incentivizes human content creation is threatened by conversational AI, which provides direct answers without needing users to click on links. If human-created content dries up, we risk living in a world of AI-generated information, a world where diverse voices are lost.
This leads to another major peril: cultural homogenization. Current large language models are overwhelmingly trained on English-language data. When you ask a question, the answer is framed from a Western, and specifically American, perspective. Unlike Google Search, which personalizes results based on location (a search for “NTU” in Taiwan gets you National Taiwan University), language models give a single, standardized answer to the world. This is a subtle but powerful force for ideological and cultural unification, and it risks erasing the unique cultural contexts of the vast majority of the world’s languages and peoples.
A New Era for Education and Work
AI is already reshaping the professional world. A Harvard Business Review study found that the emergence of ChatGPT led to a 21% reduction in outsourced jobs in the U.S., particularly in writing and coding. This isn’t necessarily about job loss; it’s about a 20% increase in the productivity of existing employees who can now use AI to perform tasks they once outsourced. Even at Google, 25% of code is now written with AI assistance.
This revolution is coming for the classroom. You are already using these tools for your assignments, and that’s not something to be ashamed of. However, our education system, built on memorization and standardized testing, is becoming obsolete. The future of education lies not in finding answers, but in asking the right questions—a shift towards a Socratic, debate-driven model of learning.
The dilemma we face is similar to the one Doraemon presents: is AI a tool that empowers us, or a crutch that makes us lazy? For young children still developing fundamental skills, over-reliance is a real danger. For university students, the answer is clearer. Top institutions like Harvard are adapting by saying, “Use AI as much as you want, but prove your understanding in an oral exam.” The goal is not to ban the tool, but to ensure that the knowledge becomes your own.
Taiwan’s Strategic Crossroads
For Taiwan, this technological shift coincides with our demographic one. The recent global expansion of TSMC should not be seen as a hollowing out of Taiwan. It should be seen as “Taiwan becoming the world’s Taiwan.”
With only 130,000 births a year, we can no longer sustain a world-leading industry with domestic talent alone. TSMC must recruit from the global talent pool of billions. Being pushed to internationalize forces us to transition from a manufacturing-based economy, which relies on population, to a knowledge-based one, like Israel, Finland, or Ireland. These small, advanced nations thrive by creating high-value intellectual property and integrating with the global market. AI is the perfect catalyst for this transformation.
Your Greatest Asset: The Skills AI Can’t Replicate
So, what should you focus on during your time here? A UNESCO official put it perfectly: “Technical skills may no longer be important, but soft skills… such as the ability to ask the machine the right questions, are crucial.”
The skills that will define your career are the ones AI cannot replicate:
- Critical Thinking and Discernment: The ability to question, to sense falsehoods, and to synthesize information with wisdom.
- Communication and Expression: The art of framing questions, articulating ideas, and inspiring action.
- Empathy and Collaboration: The capacity to understand others’ needs, build consensus, and lead teams.
- Negotiation and Leadership: The skill of finding common ground and guiding people toward a shared goal.
Our education system has long over-emphasized academic knowledge (智育). Now, we must focus on these other pillars. Move from being a “T-shaped” person with one deep specialty to a “Pi-shaped” (π) person with expertise in at least two fields, giving you greater resilience and adaptability.
To master AI, treat it like a collaborator. Learn to ask better questions. Don’t just ask, “How do I prepare for a marathon?” Instead, provide context and define roles: “You are an expert marathon coach. I am a complete beginner with 30 minutes to train each day. Create a six-month training plan for me to complete my first marathon.”
The difference between plagiarism and creation lies in effort. One prompt and one copy-pasted answer is plagiarism. An iterative conversation with 20 or 50 prompts, where you guide, refine, and challenge the AI, is a creative process.
The AI era is here. It can be intimidating, but it is also full of opportunity. Don’t fear making mistakes. Use these tools to augment your intelligence, not replace it. Your greatest challenge and your greatest opportunity is to cultivate the deep, uniquely human wisdom that AI can never possess. Focus on that, and you will not only survive but thrive in the age to come.