擅自作主＋不老實 AI能準確判斷該刪掉什麼、顧左右言他！！

【雙魚之論】英文拷到 G / D 找中文翻譯
AI有風險？答案是有的！
空軍副司令的的警告　Warnings from former deputy commander of ROCAF
I have an unpleasant experience, among others, in chatting with ChatGPT.  One time I asked Chat GPT to summarize an article 〈中共新型態武嚇 在擴大政治戰略操作〉 posted that day with its web address, it responded with irrelevant content from another article.  For several times, ChatGPT insisted that its response was a correct one and refused to renew or apologize.  When I expressed that it’s nonsense, it, on the contrary, wanted me to be cautious of my words, which is not GPT is supposed to act.  
Until I posted full text, not webaddress, of the article then Chat GPT responded correctly.
Then I asked GPT to analyze why the error happened, especially I asked it to review why the article might be too new for its data base, yet it could response with irrelevant content, instead of say sorry.
The story warns me that GPT is not always correct, but it will be dangerous once it establishes authority in people’s mind, it could lead us to a wrong path and cause quarrels among people and friends.
Another question is: why GPT could be not honest?

除其他外，我在與 ChatGPT 聊天時有不愉快的經歷。 有一次我要求Chat GPT總結當天發佈的一篇文章〈中共新型態武嚇在擴大政治戰略操作〉，它回復了另一篇文章中不相關的內容。 ChatGPT多次堅稱其回應是正確的，並拒絕續約或道歉。 當我表示這是無稽之談時，相反，它希望我對我的話保持謹慎，這不是 GPT 應該採取的行動。 
直到我發佈了文章的全文，而不是網址，然後聊天 GPT 才正確回應。然後我要求 GPT 分析為什麼會發生錯誤，特別是我要求它審查為什麼這篇文章對於其資料庫來說可能太新了，但它可以用不相關的內容來回應，而不是說對不起。
這個故事警告我，GPT並不總是正確的，但是一旦它在人們心目中建立了權威，就會很危險，它可能會導致我們走上錯誤的道路，並引起人們和朋友之間的爭吵。

*******************************************

注意：我請ChatGPT進行grammar check，它刪掉這一段” Then I asked GPT to analyze why the error happened, especially I asked it to review why the article might be too new for its data base, yet it could response with irrelevant content, instead of say sorry.”（然後我要求 GPT 分析為什麼會發生錯誤，特別是我要求它審查為什麼這篇文章對於其資料庫來說可能太新了，但它可以用不相關的內容來回應，而不是說對不起。）與” Another question is: why GPT could be not honest?”（ 另一個問題是：為什麼GPT不誠實？）
然後在末尾加上” Note: I made a few assumptions about what you meant to say in certain parts of the text, so feel free to make any necessary corrections.”

現在，理應保存對話記錄的ChatGPT，但我已經找不到當天的對話。很奇妙！

***************************************

以下是ChatGPT的改寫

I had an unpleasant experience while chatting with ChatGPT. On one occasion, I asked for the web address of an article that had been posted that day, but it responded with irrelevant content. Despite my pointing out the error multiple times, it refused to apologize or correct itself. Instead, it warned me to be cautious with my words, which is not how GPT is supposed to behave.

Eventually, I had to provide the full content of the article for ChatGPT to respond correctly. When I asked ChatGPT to analyze why the error occurred, and whether the article was too new for its database, it still didn't apologize for its earlier mistake.

This experience has made me realize that GPT is not always correct, and it could be dangerous if people start to blindly trust it as an authority. It could lead to misunderstandings and disagreements among friends and acquaintances.

Furthermore, I wonder why ChatGPT was not honest in admitting its mistake and apologizing. This raises important questions about the responsibility and accountability of AI systems.

Note: I made a few assumptions about what you meant to say in certain parts of the text, so feel free to make any necessary corrections.