ChatGPT’s Chinese Responses Frustrate Native Speakers

ChatGPT’s conversational quirks in English can be grating, but its habits in Chinese are even more jarring. A Wired investigation examined how the AI handles Chinese, the world’s most spoken language with over 1.1 billion native speakers, according to the Language School at Middlebury College.

Common AI Phrases That Annoy Chinese Users

The AI frequently defaults to phrases that sound unnatural or overly dramatic to native speakers. One recurring response is:

「我会稳稳地接住你」
Literal translation: “I will catch you steadily.”
More flowery alternative: “I’ll hold you steadily through whatever comes.”

While the sentiment may seem supportive, Chinese speakers find it irritating. The phrase has become so ubiquitous that it’s now a meme, with some depicting ChatGPT as a giant inflatable airbag ready to “catch” users mid-fall.

Another overused phrase is:

「砍一刀」
Meaning: “Help me cut it once” or “slash the price.”

This line mimics ad copy from Chinese e-commerce platform Pinduoduo, further highlighting the AI’s unnatural phrasing.

Why Does ChatGPT Sound So Awkward in Chinese?

The issue may stem from a phenomenon called “mode collapse”, a bias in large language models (LLMs) caused by human data annotators. During training, annotators unknowingly favor familiar phrases over more natural-sounding ones, reinforcing repetitive responses.

Once an LLM like ChatGPT is trained, correcting these biases is difficult. Developers can encourage certain responses, but enforcing variety and natural phrasing remains a challenge.

Max Spero, cofounder and CEO of AI-writing detector Pangram:
“We don’t know how to say: ‘this is good writing, but if we do this good writing thing 10 times, then it’s no longer good writing.’”

Shared Frustration: English and Chinese Speakers Agree on ChatGPT’s Flaws

Despite language barriers, English and Chinese speakers share a common frustration: ChatGPT’s inane, repetitive babble. While the AI excels in some areas, its unnatural phrasing in Chinese highlights the broader challenges of training AI to sound human.

For more on ChatGPT’s limitations, read: Even After Two Massacres, OpenAI Still Hasn’t Stopped ChatGPT From Helping Plan School Shootings

The post ChatGPT Is Saying VWeird Things in Chinese first appeared on Futurism.

Source: Futurism