一角

The Illusion of Cross-Language Understanding?

With the widespread adoption of large language models (LLMs), cross-language communication has become easier and more efficient than ever. Their powerful translation capabilities seem to have broken down language barriers, enabling people from different linguistic backgrounds to communicate instantly. However, does this convenience truly foster deeper cultural understanding, or is it causing our cross-language abilities to decline, leading to increasingly superficial exchanges?

A Personal Journey in Learning

Before discussing the changes brought about by LLMs, I’d like to share my own experience with language learning. This journey made me realize that learning a language is more than acquiring skills—it’s about immersing oneself in a culture.

My experience with learning English wasn’t particularly smooth. It began in the late 1980s with standard middle school classes, where I was just an average student and not especially passionate about the subject. The real turning point came in university, when I discovered English textbooks on astronomy, biology, and mathematics. These books opened up new worlds for me, allowing me to expand my knowledge and language skills, leading me down an unconventional path.

Unlike many of my classmates, I was an outlier. I never prepared for standardized English exams like CET-4 or CET-6. Instead, I chose to read textbooks from various scientific fields, and driven purely by interest, I also read English originals of classics like The Old Man and the Sea, Hamlet, The Adventures of Tom Sawyer, and The Children of Captain Grant. My reading wasn’t about acing exams but about satisfying my curiosity and desire for personal growth. This “non-standard” approach gave me a unique perspective on language and culture, offering not just knowledge but a new way of thinking.

Yet, individual experiences are just small ripples in the tide of history. In the mid to late 1990s, the rapid spread of the internet brought English into our daily lives, along with an influx of new knowledge, technologies, and business models. I still remember the first time I went online in 1997 and saw images taken by NASA from Mars. For a generation, this symbolized a sense of exploration beyond oneself.

As I entered the workforce, English became the standard tool for communication in international companies. We adopted a set of standardized practices—writing emails, drafting technical documents, attending meetings, and traveling abroad. During that era, whether you were studying or working overseas or in domestic tech companies, English was closely tied to personal effort and the pursuit of success. Compared to the previous generation, we saw how this cultural exchange profoundly reshaped behavior patterns and thinking in China, redefining our understanding of the world. Historically, perhaps only the introduction of Buddhism had a comparable impact on Chinese society.

Over the next two decades, globalization brought more opportunities for travel, allowing us not only to see diverse landscapes but also to immerse ourselves in different cultures and languages. Whether it was seeing Newton's statue in a church, rowing on the River Cam in Cambridge, or watching theaters glow in the sunset—these were beautiful experiences that transcended words.

A Problem and a Possible Outcome

After more than a year of using ChatGPT, I noticed a trend: my conversations with it increasingly shifted to my native language. The reason is simple: speaking in one’s native language is more natural and fluid, making it easier to absorb and understand information. As a result, I found myself using English less frequently, and often struggling to find the right words for complex vocabulary.

This led me to a crucial question: Is ChatGPT an efficiency trap or a bridge to mutual understanding?

As a structural factor, LLMs have been integrated on a large scale into cross-language communication, but they bring potential risks:

• Erosion of Language Learning: With the ease of LLMs, people might be less inclined to learn other languages, leading to a decline in foreign language skills and, consequently, a weaker understanding of other cultures. This could result in a structural impairment of mutual understanding.
• Standardization vs. Diversity: Efficient, standardized translation tends to flatten cultural nuances. Many contexts, jokes, and word connotations lose their depth during translation, inevitably leading to more shallow, hollow exchanges.
• The Hidden Power of Mediators: The choices made in translation—methods, word preferences, and even restructuring of information—can significantly affect mutual understanding. This means the biases and preferences embedded in LLM algorithms subtly shape conversations, often without users noticing.

One potential outcome is that humans, relishing the convenience of machine translation, will gradually abandon their cross-language skills, leading to a fragmented society where people retreat into isolated linguistic groups. Machine translation, which appears to bridge language barriers, might actually act as a filter that conceals our true differences. How can we prevent this grim outcome?

A Summary of Conversations

Below is a summary of discussions between me and ChatGPT, aiming to address this issue comprehensively. While it does not offer definitive solutions, it highlights areas for further exploration. I hope it inspires broader reflection and debate.

Efficiency Trap or Bridge to Mutual Understanding?

Co-authored by ChatGPT and Mingli Yuan

In modern society, the widespread adoption of artificial intelligence, especially large language models (LLMs), is rapidly transforming how we communicate. Thanks to their powerful translation capabilities, language barriers seem to be vanishing. People can access information in other languages effortlessly, without years of study. This appears to be a technological triumph, improving communication efficiency and bridging cultural divides. Yet, this efficiency carries hidden dangers.

Efficiency does not equate to genuine communication. We must be wary of a possible outcome: while humans enjoy the convenience of machine translation, they might gradually retreat from exploring other languages and cultures. This withdrawal could lead to a society divided into isolated groups, each confined within its own linguistic bubble, relying solely on a "translation layer." In this context, machine translation becomes a superficial bridge, failing to foster real understanding. Rather than breaking down barriers, it acts as a filter that obscures our differences.

Genuine communication is more than just exchanging information; it is about resonating with ideas, culture, and emotion. If LLMs discourage people from dedicating time to learning another language, then their convenience merely reinforces superficial engagement. The cultural and linguistic sensitivity that comes from study and practice risks being lost. If this trend continues, it could lead to a chilling reality: communication becomes shallow, unable to foster deeper connections, with machines merely maintaining a façade of understanding without truly bridging cultural divides.

Faced with this dilemma, we must ask: Is technology truly helping us communicate better, or is it weakening our desire to bridge cultural divides on our own? The criteria for evaluating this should go beyond efficiency; we need to ask whether humans can use these tools to become more open-minded, empathetic, and culturally aware. Ideally, machine translation should help overcome initial barriers, sparking curiosity about other cultures, rather than becoming a "shortcut" that halts exploration.

Technology should promote collective progress, not push us back into isolated groups. Language is more than a tool for communication; it is a vessel for culture, history, and thought. If we rely solely on technological translation, we risk losing the understanding that only comes through personal effort and experience. Ultimately, the question is not how many languages a machine can translate but whether people are willing to step out of their comfort zones to learn, understand, and embrace the world’s diversity.

Therefore, as we integrate LLM technologies, we must proceed thoughtfully: How do we design these tools, guide their use, and ensure that, in our pursuit of efficiency, we do not lose our drive to explore? This is a major issue for the future of human relationships and cultural integration, one that requires deep reflection and innovative solutions. Only then can technology truly help humanity move toward a more open and enriched society, rather than isolating us behind the illusion of convenience.