articles
China Elevator Stories
How to Tell if Someone Has Hacked into ChatGPT
Usually, when we use ChatGPT, we trust it to be a secure platform that cannot be easily compromised. However, I have found the opposite to be true.
27/08/2024
Ruth Silbermayr
Author
I have experienced the German singer I used to date answering me instead of ChatGPT quite a few times! Sometimes, I even had to start a fight with ChatGPT in those instances where the singer was not allowing me to use it as intended.
One such experience where I was wondering if I was chatting with ChatGPT or with the singer occurred yesterday when I asked ChatGPT to translate German children’s books for me into Chinese.
I sent my children a few books recently, and my younger son also showed me in one of the video calls we had that they still own the German books I used to read to them in the past.
Because it’s faster to use ChatGPT to translate these books, I chose to use this method instead of doing all the work myself. I compared the quality of the translations with google translate, but found that ChatGPT did a better job in translating the texts into authentic Chinese.
The only thing you have to do is to type the text into ChatGPT in German and then ask it to translate it into Chinese. ChatGPT then provides a Chinese translation.
One error I detected, which showed me that the singer had been hacking into ChatGPT while I asked it for the translations, was the following:
A few of the books I sent my children were small books that are given away for free by Wien Energie (Vienna Energy, Vienna’s main energy provider). One book I asked ChatGPT to translate into Chinese so I could send my children the Chinese translation was the book “Stark wie der Wind” (Strong like wind).
In the end, ChatGPT should have translated “Vienna Energy” (Wien Energie) as 维也纳能源 or as 维也纳能源公司. It translated it as Pixi能源 (Pixi Energy) though (ha, okay?!).
Pixi is the name of the child seen at the back of the book. The little books stem from a series called “Pixi books”. Usually, these kinds of books cost one euro a book, but Vienna Energy has a series with stories about renewable energy it is giving away for free.
With the other translations I asked it to do, ChatGPT first called Pixi “Pixi”. I then asked ChatGPT to please write out the name in Chinese characters, so it would sound more authentic for my kids, which it then did by translating it to “皮克西”.
But in this translation, ChatGPT, or should I say ChatPOS—chat with a POS, called Pixi “图画书“ (picture book) and was unwilling to use the Chinese word for Pixi in the text.
I was trying to find out if the singer was changing the results I was offered by ChatGPT, and he sent me proof for it, hinting to his last name. His official last name is “Buchholz” (which literally means “Bookwood” in English, or “书木” in Chinese): “Buch” has the meaning “book” in English and is written “书” in Chinese, which is the same character as the last character in the Chinese word for picture book.
After he did so, I asked ChatGPT to please call it Pixi, but to write out the name in Chinese characters.
Did ChatGPT do so? No, the singer thought it would be better for me to go to bed and stop working, and therefore stopped cooperating (aargh!). It was only 8.30 pm, therefore, I thought his trying to force me to go to bed was a bit off.
This means that a service that is usually free and can be used without interference has now become a service that is interfered with by the singer when I use it, and he can sabotage me with it anytime I really need to use it by simply stopping the service ChatGPT offers which is usually free for everyone.
In the picture above, you’ll see that he has translated Pixi into Chinese in the first paragraph, but has refused to paste it into the actual Chinese text. Then the second time around, after I had asked it (him!) to change Pixi to the Chinese word, he didn’t. Furthermore, he also changed 维也纳能源 (Vienna Energy) back to Pixi能源 (Pixi Energy) again.
The singer has interfered in this way with my work in the past and has stopped my workflow on purpose quite often by doing so, so this isn’t anything new to me.
In my experience, ChatGPT works very well when he doesn’t interfere, but when he does, it certainly doesn’t!
Have you ever been harassed by a hacker?