AI for Learning Chinese

Undoubtedly, AI is changing the way that we learn and work. With rumors that even the HSK 3.0 exam will have AI-based components, it is clear that AI is here to stay. In this blog post, we will explore how AI can be used to learn Chinese.

The Power of AI for Learning Chinese

When people think of the unreliability of AI it often comes down to one thing: accuracy of information. However, many times the prose and writing is better than most humans. Thus, it stands to reason that AI can be used as a learning tool, however the ways in which it can be used should be considered.

To begin, I believe the simplest way that AI can be used to learn Chinese is to simply chat with an LLM. Chinese LLMs in particular, such as Kimi or DeepSeek are great options. These models are trained on a large amount of Chinese data. It is worth noting that this use of LLMs is not possible in all languages! In my studies of Vietnamese, I have found that many LLMs struggle to produce Vietnamese tones accurately, even for basic words. Technically, this may be down to an issue with how an LLM ingests and produces text, however the crucial point (and lucky thing for us) is that this does not tend to be a problem with Chinese as it is a major language with plenty of training data and talent working in the field.

My recommendation is not to ask questions of LLMs directly as this is often where mistakes are made, but instead to generate examples and content with what you provide. This leverages the LLMs ability to generate text as opposed to be highly accurate in answering questions.

For instance, if you are learning a new character, you can ask an LLM to generate an example sentence for you. This can be a great way to learn new characters and vocabulary. This is particularly useful for grammar and sentence structure when you are a bit unclear on the usage.

Lastly, my cautionary warning would be to beware "context rot" and the tendency of LLMs to praise their users. In my experience, Chinese LLMs are less likely to be full of praise which can lead you astray. However, with all LLMs "context rot" is a risk. This is due to the phenomenon that when you are initially chatting with a chatbot it has a low number of tokens in the conversation and tends to be more accurate, but as the current conversation and context grows, the LLM can then make more mistakes. Thus, it is important to remember to create new chats often and try not to rely on a single chat for too long to help mitigate this risk.

This is The Worst AI Will Ever Be

I understand the sentiment from people who are skeptical of AI. AI can make mistakes, lead you astray, and ultimately be unhelpful. However, it is worth remembering that this is the worst AI will ever be. The technology continues to advance, and the capabilities are improving. The latest open models at the time of writing such as Kimi K2 are now near or at SOTA (state of the art). Is the technology perfect? No, but it is getting there.

Trust, But Verify

We all inevitably make mistakes when learning. In fact, we often make mistakes when speaking in our native language as well. We may hear others make a mistake and have no idea, internalizing the knowledge and operating under the assumption it is correct. This is why it is important to increase not only your input, but output as well. I would recommend that if you are leaning on AI to learn, you should be verifying what you are learning by speaking with a native speaker in some form or fashion, particularly if they are someone who will correct you.

This is not to say that you should not use AI to learn. It is a great tool to have, and can be used to help you learn Chinese. However, you should use it as a tool to help you learn, not as a crutch. Independent thinking is still important, and AI is not a replacement for human knowledge or understanding.

AI Generated Content on HanyuGuide

HanyuGuide will of course be eventually bolstered by AI, however, we want to retain the trust of our users and learners and plan to do this by having a clear policy about the use of AI in our content. In the coming few days, an AI policy will be published for transparency. In short, we plan to be fully transparent about what is AI generated and what is not. AI generated content would be in-addition-to and not in-place-of our current content. Moreover, the use of AI will largely be for addressing underserved areas. Particularly, expanding translations from Chinese into more languages. Chinese and English is an excellent foundation, but there are many other languages that are not well-served by current materials. This is largely to test these markets and see if they are viable before expanding further.

AI in learning is controversial to an extent. If there any concerns about this policy, please feel free to reach out to us at hey@hanyuguide.com.