[The World on the Page] In an AI World, Deep Study in Your Field Matters Even More

Professor Baron, Author of "The Future of Writing"
"AI Prioritizes Efficiency Over Accuracy"
Taming Artificial Intelligence Is Essential
The More You Know, the Better You Handle AI
Critical Literacy Requires Systematic Learning

The whole world was once again stirred by artificial intelligence. The shock of DeepSeek, developed by a small startup in China, left us in awe. It was astonishing that a group of young researchers around thirty years old produced an AI with performance comparable to that developed by Silicon Valley companies in the U.S., which spent tens of billions of won, all for less than 10 billion won. Especially remarkable was the fact that they did not use high-spec chips that the U.S. has been desperately trying to control exports of.

Clearly defined limitations rarely discourage humans. The reason we hesitate to challenge problems is that we do not know for what purpose, how far, or how much effort is needed. The emergence of DeepSeek is a good example that reminds us of the truth that known finiteness ultimately stimulates human creativity.

Of course, there have been suspicions about whether only that amount of money was spent on DeepSeek’s development, and various issues have arisen, such as the revelation that some responses from DeepSeek were contaminated with China-related content. However, the undeniable fact remains that DeepSeek presented a new method to develop AI at a revolutionary low cost.

What the DeepSeek model truly means for our lives is that the development of AI in various fields will accelerate significantly. Like ChatGPT, the reduced development cost of DeepSeek, which utilizes large language models, is thanks to expert modules. While traditional AI models answer questions by searching through entire books and learning from them, DeepSeek first consults experts and then, if insufficient, searches more books to learn and respond. Since most companies only need to excel in their specific fields, many companies can now embark on developing the AI they need. It feels like an era is rapidly approaching where AI coexists and collaborates in every field.

Additionally, intellectual abilities traditionally acquired through university or graduate education are at risk of being replaced more quickly. Tasks such as translating documents, reviewing loan applications, collecting laws and precedents and organizing issues in court cases, or analyzing accounting data to find optimal tax-saving methods are examples. There is much debate on how to respond.

As many experts warn, first remember that AI’s answers suffer seriously from hallucinations (errors), so it is essential to have critical literacy to handle them. Whether ChatGPT or DeepSeek, current language learning models cannot fundamentally solve the error problem beyond a certain level. Moreover, it is common for them to be contaminated with biases and harsh remarks related to race, ethnicity, gender, or status. For example, AI used in video chat job interviews often shows a ‘bookshelf bias.’ If a bookshelf full of books is visible in the background, the AI tends to take the applicant’s statements more seriously.

As Naomi Baron, Professor Emerita at American University in the U.S., revealed in The Future of Writing (BookTrigger), AI using large language models fundamentally “does not understand how the world actually works or how humans grasp such workings.” It merely collects sentences on a large scale, arranges them according to sophisticated algorithms, and automatically generates sentences. These answers generally prioritize efficiency over accuracy.

For this reason, Professor Baron says it is important to properly tame AI to suit us. If we fully entrust writing, that is, answering questions, to AI, our ability to reflect on ourselves weakens. We write to “convey meaning, share hope, find ourselves, and understand the world.” AI has no interest in this. When asked a question, it merely rummages through everything written by humans, edits it, and throws an appropriate answer on the screen.

Moreover, in the process of proofreading, editing, and revising the drafts we first write?our provisional answers?we often gain “greater wisdom by contemplating, criticizing, and improving our thinking.” Furthermore, before showing our writing to others, we verify its validity and truthfulness, consider whether it can persuade others, and carefully polish words and sentences to avoid hurting others’ feelings. This process fosters the ability to face community realities and reflect on human relationships. If we leave this work to AI, our ability to think about these issues may weaken.

However, even if the answers that pour out instantly upon asking contain random errors and garbage-like biases, the whole is not worthless, of course. For example, spell check functions when writing are still imperfect and often make mistakes. But I often tell students to run an automatic check once before submitting assignments because it helps beginners quite a bit. Of course, experts use this function better. They can accurately judge which sentences and words are correct or incorrect, allowing them to immediately correct unconscious mistakes while writing.

Recently, many in the publishing industry are considering working by first using AI to perform a primary translation of documents, then having editors proficient in Korean and foreign languages polish the text while comparing it with the original. This speeds up the work and may reduce mistranslations. The practice of verifying uncertain knowledge by first asking AI is also spreading, much like Go players reviewing optimal moves with AI. Such collaboration between humans with expertise and AI will undoubtedly greatly change our daily landscape.

It is extremely rare for any machine to realize the correct way to use itself automatically. AI also requires learning and mastering how to use it properly. Everything is seen well as much as one knows and handled well as much as one studies. In a world overflowing with AI, I believe humans who deeply study their fields and master the details become even more important.

Baron says, “Human writing sharpens the human mind and is a magical sword that connects us to others. No matter how efficient AI is, it is our responsibility to keep that sword shining.” Critical literacy does not arise without systematically learned knowledge, training to place knowledge in a broader context and reflect on it, and repeated practice of writing and refining sentences. This fact does not change even in the AI era.

Jang Eun-su, Publishing Culture Critic