AI and Human Writing
AI and Human Writing
AI 與人類寫作
Introduction
AI can now write texts. It is hard to know if a human or a computer wrote a story.
現在 AI 已經可以寫文章。很難分辨故事是由人類還是電腦寫的。
Main Body
Many people think they can find AI writing. They are often wrong. Professor Claire Hardaker says people are only right 60% of the time. AI learns from humans, so they write in the same way.
許多人認為他們能發現 AI 的寫作。但他們通常是錯的。Claire Hardaker 教授表示，人們僅在 60% 的時間裡判斷正確。AI 向人類學習，因此他們的寫作方式相同。
Some writers use AI and then lie about it. Some people use special tools to find AI text on websites. These tools are not perfect. They often miss the AI text.
有些作者使用 AI 之後會對此撒謊。有些人使用特殊工具來尋找網站上的 AI 文本。這些工具並不完美，經常會遺漏 AI 文本。
AI does not have a body or real feelings. Professor Peter Stockwell says AI only uses old information. It cannot create new or strange ideas like humans can. AI makes all writing sound the same.
AI 沒有身體或真實的情感。Peter Stockwell 教授表示 AI 僅使用舊資訊。它無法像人類一樣創造新穎或奇特的想法。AI 讓所有寫作聽起來都一樣。
Conclusion
We cannot always tell the difference between humans and AI. But AI cannot be truly creative because it is not alive.
我們無法總是分辨人類與 AI 的區別。但 AI 無法真正具有創意，因為它不是生物。
Vocabulary Learning
🧩 The Power of "CAN" and "CANNOT"
In this text, we see a very important word for A2 students: Can. We use it to talk about abilities (things we are able to do).
How it works:
- Positive: Subject + can + action word.
- Negative: Subject + cannot + action word.
Examples from the text:
- AI can write texts. → (It has the ability)
- AI cannot create new ideas. → (It does not have the ability)
- We cannot always tell the difference. → (It is impossible for us)
💡 Simple Rule: Notice that we do not add "s" or "ing" to the action word after can.
❌ AI can writes (Wrong) ✅ AI can write (Right)
Quick Comparison:
- Human → Can be creative ✅
- AI → Cannot be truly creative ❌
Vocabulary Learning
The Rise of Large Language Models and the Problem of Verifying Human Writing
大型語言模型的興起與驗證人類寫作的問題
Introduction
The use of generative artificial intelligence in writing has caused a widespread problem: it is now very difficult to tell the difference between AI-generated text and writing created by humans.
在寫作中使用生成式人工智慧引起了一個普遍的問題：現在很難區分 AI 生成的文本與人類創作的寫作。
Main Body
Currently, there is a big difference between what people believe and the actual facts regarding AI detection. Professor Claire Hardaker's research shows that humans are only about 60% accurate when trying to identify AI text. This suggests that looking for simple patterns, such as specific punctuation, is not a reliable method because AI is trained on human writing. Consequently, human writers and AI models often end up using the same styles, making them hard to distinguish.
目前，人們對 AI 檢測的認知與實際事實之間存在巨大差異。Claire Hardaker 教授的研究顯示，人類在嘗試辨識 AI 文本時，準確度僅約 60%。這表明，尋找簡單的模式（例如特定的標點符號）並非可靠的方法，因為 AI 是根據人類的寫作進行訓練的。因此，人類作者與 AI 模型經常使用相同的風格，導致兩者難以區分。
Many professional and creative fields have seen an increase in accusations of using AI. In the literary world, some authors have had to withdraw their work or apologize for fake citations created by AI. Similarly, in the fanfiction community on the AO3 platform, users have created technical tools to find hidden code left by AI models like Claude. However, these tools only work for direct copy-pasting and cannot detect text that has been edited, which leads to many mistakes.
許多專業與創意領域中，關於使用 AI 的指控有所增加。在文學界，部分作者不得不撤回作品，或為 AI 產生的虛假引用道歉。同樣地，在 AO3 平台的同人小說社群中，使用者開發了技術工具來尋找 Claude 等 AI 模型留下的隱藏代碼。然而，這些工具僅對直接複製貼上有效，無法檢測經過編輯的文本，這導致了許多錯誤。
Experts argue that while AI is good at basic grammar, it lacks the real-life experience and emotions needed for true creativity. Professor Peter Stockwell emphasizes that AI is naturally conservative because it only uses existing data. Therefore, it cannot create the original or rebellious styles that humans produce. Furthermore, this leads to 'cultural ghosting,' where language becomes standardized and loses its global diversity.
專家認為，雖然 AI 擅長基礎語法，但缺乏真正創意所需的現實生活經驗與情感。Peter Stockwell 教授強調，AI 本質上是保守的，因為它僅使用現有數據。因此，它無法創造出人類所產生的原創或反叛風格。此外，這會導致「文化鬼影」（cultural ghosting），使語言變得標準化並失去全球多樣性。
Conclusion
Because we cannot reliably tell human and AI text apart, a climate of suspicion has developed. Nevertheless, the fact that AI lacks real consciousness means it still cannot achieve genuine literary innovation.
由於我們無法可靠地分辨人類與 AI 的文本，一種懷疑的氛圍隨之而來。儘管如此，AI 缺乏真實意識這一事實，意味著它仍然無法實現真正的文學創新。
Vocabulary Learning
⚡ The 'Logic-Link' Shift
To move from A2 to B2, you must stop using only simple words like and, but, and because. B2 speakers use Connectors of Result and Contrast to make their arguments sound professional and academic.
🧩 The 'Result' Chain
In the text, we see a sophisticated way to show cause and effect. Instead of saying "AI is trained on human writing, so it looks human," the author uses:
*"...AI is trained on human writing. Consequently, human writers and AI models often end up using the same styles..."
The Upgrade:
- A2: So... B2: Consequently / Therefore / As a result
Example: I didn't study the verbs. Consequently, I failed the exam.
⚖️ The 'Contrast' Pivot
Notice how the author introduces a surprising fact or a contradiction. They don't just use but.
*"...a climate of suspicion has developed. Nevertheless, the fact that AI lacks real consciousness..."
The Upgrade:
- A2: But... B2: Nevertheless / However / Despite this
Example: The hotel was very expensive. Nevertheless, the service was terrible.
🚀 Pro-Tip: The Punctuation Secret
B2 learners often forget that these 'power words' usually need a period (.) or a semicolon (;) before them and a comma (,) after them.
Wrong: I am tired but I will work. B2 Style: I am exhausted. However, I will finish the project tonight.
Vocabulary Learning
The Proliferation of Large Language Models and the Resultant Crisis of Linguistic Authentication
大型語言模型的普及與由此引起的語言認證危機
Introduction
The integration of generative artificial intelligence into textual production has precipitated a systemic difficulty in distinguishing synthetic outputs from human authorship.
生成式人工智慧融入文本創作，導致目前在區分合成輸出與人類創作之間存在系統性困難。
Main Body
The current landscape of textual authentication is characterized by a significant discrepancy between public perception and empirical reality. Forensic linguistic analysis conducted by Professor Claire Hardaker indicates that human accuracy in identifying AI-generated text is approximately 60%, suggesting that reliance on simplistic stylistic rubrics—such as the 'rule of three' or specific punctuation patterns—is fundamentally flawed. These markers are often inherent to human writing, as large language models (LLMs) are trained on human corpora, creating a recursive linguistic loop where human writers are simultaneously influenced by AI stylistic norms.
目前的文本認證格局呈現出大眾認知與實證現實之間的顯著差異。Claire Hardaker 教授進行的法庭語言學分析指出，人類辨識 AI 生成文本的準確率僅約 60%，這表明依賴簡單的風格準則（如「三條規則」或特定的標點符號模式）基本上是有缺陷的。這些標記通常是人類寫作本身固有的，因為大型語言模型 (LLM) 是基於人類語料庫訓練的，從而創造出一個遞迴的語言迴路，使得人類作者同時受到 AI 風格規範的影響。
Institutional and creative sectors have experienced a surge in accusations of synthetic authorship. In the literary domain, this has manifested in the withdrawal of publications and public apologies for 'hallucinated' citations. Within the fanfiction community, specifically on the Archive of Our Own (AO3) platform, users have deployed technical 'skins' to detect specific HTML artifacts injected by Anthropic’s Claude. While these tools can identify direct copy-paste actions, they fail to detect text processed through intermediary editors, thereby introducing a high risk of false negatives and an inability to quantify the extent of AI assistance.
機構與創意領域對合成創作的指控大幅增加。在文學領域，這表現為撤回出版物以及為「幻覺」引用而公開道歉。在同人小說社群，特別是在 Archive of Our Own (AO3) 平台上，用戶部署了技術「皮膚」以偵測由 Anthropic 的 Claude 注入的特定 HTML 痕跡。雖然這些工具可以識別直接複製貼上的行為，但無法偵測經過中間編輯器處理的文本，從而引入了高風險的偽陰性，且無法量化 AI 輔助的程度。
Academic perspectives suggest that while LLMs excel at lower-level syntactic structures, they lack the 'embodied' experience and limbic system necessary for high-level narrative innovation. Professor Peter Stockwell posits that AI is inherently conservative, as it operates on existing data and cannot replicate the subversive or avant-garde impulses driven by human social and physical existence. Consequently, the 'flattening' of language toward an Anglo-American standard—termed 'cultural ghosting'—represents a significant institutional shift in global linguistic diversity.
學術觀點認為，雖然 LLM 在低階句法結構方面表現卓越，但缺乏高階敘事創新所需的「體現」經驗與邊緣系統。Peter Stockwell 教授認為 AI 本質上是保守的，因為它運作於現有數據，無法複製由人類社會與生理存在所驅動的顛覆性或前衛衝動。因此，語言向英美標準的「扁平化」—— termed 為「文化鬼影」——代表了全球語言多樣性在制度上的重大轉移。
Conclusion
The inability to reliably differentiate between human and synthetic text has fostered a climate of suspicion, though the fundamental lack of sentient experience in AI continues to limit its capacity for genuine literary innovation.
由於無法可靠地區分人類與合成文本，導致了一種懷疑氛圍，儘管 AI 根本缺乏感知經驗，仍持續限制其進行真正文學創新的能力。
Vocabulary Learning
The Architecture of Nominalization & Lexical Density
To transition from B2 (where communication is fluid) to C2 (where communication is authoritative), one must master the art of Nominalization. This is the linguistic process of turning verbs or adjectives into nouns to create a denser, more academic tone that shifts the focus from actors to phenomena.
⚡ The C2 Pivot: From Process to Concept
Observe how the author avoids simple subject-verb-object structures to create a sense of objective detachment.
- B2 Level (Action-oriented): AI is being integrated into how we write, and this has made it hard to tell if a human or a machine wrote the text.
- C2 Level (Nominalized): "The integration of generative artificial intelligence into textual production has precipitated a systemic difficulty in distinguishing synthetic outputs from human authorship."
The Linguistic Shift:
- "Integration" (Noun) replaces "integrating" (Verb).
- "Textual production" (Noun phrase) replaces "how we write" (Clause).
- "Systemic difficulty" (Noun phrase) replaces "made it hard" (Adjective phrase).
🔍 Dissecting "Recursive Linguistic Loops"
C2 mastery involves using conceptual metaphors embedded in high-level vocabulary. The phrase "recursive linguistic loop" is a masterstroke of precision. It doesn't just mean "a circle"; it implies a mathematical/computational process where the output of one stage becomes the input for the next.
Key C2 Collocations for your arsenal:
- Precipitate a difficulty (instead of "cause a problem")
- Empirical reality (instead of "the facts")
- Subversive impulses (instead of "wanting to break rules")
- Quantify the extent (instead of "measure how much")
🚩 The Danger of "Flattening"
Note the use of metaphorical verbs in an academic context: "the 'flattening' of language." In B2, you use "flatten" for a pancake; in C2, you use it to describe the loss of nuance and diversity in a global system. This is the hallmark of an expert user: using concrete physical imagery to describe abstract sociological shifts (Cultural Ghosting).