留子们用火星文写避雷帖,AI竟看懂了?我们实测:几乎全军覆没

留子们在写一种很新的差评。
由于全球点评打卡平台,都有个自动翻译的功能,为了防止被店家骚扰,也为了给同胞避雷,身在海外的国人们,想出各种奇招优雅给差评。


例如,打 5 分忽悠老板,再用中文加密:
图片
用中文谐音写了一堆差评,然后在结尾给个虚假好评,商家拿翻译器一翻,只看到一个早餐不错。
图片
还有人费老大劲儿,用米粒摆出「报吃」的汉字,中国人一看就懂:
图片
更有甚者,重启上古火星文,直接干翻各类翻译器。
图片
但是,最近有网友发现,这段如同计算机乱码的火星文差评,AI 竟然也能读懂啦。


I’m analyzing an encrypted hotel review with a mix of Chinese, Russian, phonetic symbols, and Japanese hiragana. The text might be in “Martian language” using homophones and similar characters. Let’s break it down.

OK, let’s break down the title “卟崾莱住” into “不要来住” by matching similar characters. This tells me it’s about avoiding a stay, which seems interesting.

I’m matching transliterated Chinese characters to Cyrillic ones, tweaking only the vowel sounds. Each line translates smoothly, confirming the title’s meaning.

I’m translating and interpreting Chinese text, thinking through the use of homophones and characters to form meaningful sentences, like identifying the meaning of “笩洃姑娘” as “灰姑娘”.

标题:“不要来住”
正文:“中国人不骗中国人!墙壁和纸一样薄,这么贵的价格房间竟然在阁楼,难道我是当灰姑娘?洗澡水又少,厕所的灯光也昏暗,哈利波特都比我住的呢。避雷!都不许来住!!!”

I’m analyzing an encrypted hotel review with a mix of Chinese, Russian, phonetic symbols, and Japanese hiragana. The text might be in “Martian language” using homophones and similar characters. Let’s break it down.

I went to eat at an old-fashioned Beijing cuisine restaurant during the New Year, and the plaque was shining and the restaurant was covered with Guo Moruo’s inscription. After eating, I was extremely disappointed, and my childhood happy memories were all shattered… A friend described it as the kind of taste that even Guo Moruo would jump out of his coffin when he knew how bad it was.

(Friendly reminder: Using Tencent’s “Martian Text Translator” from “Help with a Little” app)

Link:

https://tool.browser.qq.com/toMars.html

Except for Byte Dough Pack’s accurate translation of Martian text, other domestic and foreign large models all failed.

Taking out “Book Boon,” the translation methods of other large models for “豪华若” are all over the place.


These are technical analysis and translation results on how various AI models interpret encrypted text written in Martian language. Despite challenges, these endeavors underline the evolving capabilities and limitations of AI-powered language processing tools.

感谢阅读!如果您对AI的更多资讯感兴趣,可以查看更多AI文章:GPTNB

留子们用火星文写避雷帖,AI竟看懂了?我们实测:几乎全军覆没

https://www.gptnb.com/2024/09/21/2024-09-20-auto5-NAyIqh/

作者

ByteAILab

发布于

2024-09-21

更新于

2025-03-21

许可协议