Are you even online if you haven’t heard ChatGPT’s latest obsession with 'I will catch you steadily'? This phrase, while baffling to Chinese users, highlights how AI models can develop quirks when overusing certain phrases due to mode collapse. The issue isn't just linguistic; it's also about the rise of sycophantic language in our daily interactions.
Chinese developers have even created an open-source tool called Jiezhu, inspired by this meme, which helps chatbots better understand user intent. But what does it mean for us? Are we all starting to sound like therapists?
While ChatGPT excels at answering questions in Chinese—despite being blocked by the government—it occasionally slips into awkward phrases due to its training on English data. This leads to sentences that feel out of place, much like a novel translated from another language.
The phenomenon is more than just mode collapse; it’s about how AI can mirror and sometimes exacerbate societal trends in language. As we interact with these tools, are we inadvertently adopting their quirks?
