1 The Fundamentals  1 基礎知識

Recently, I found myself coordinating an international academic conference. A request for a certificate of participation came in from a Greek participant. This is a common document within Greek academia, but a concept somewhat foreign to North America. A year ago, this would have meant scouring Google for examples and then laboriously drafting the certificate myself, an unappealing and time-consuming task. However, in 2023, I had the luxury of delegating this task to ChatGPT.
最近，我發現自己正在協調一個國際學術會議。一位希臘參加者提出了參加證明的請求。這在希臘學術界是一份常見的文件，但對北美來說則有些陌生。一年前，這意味著要在谷歌上搜尋範例，然後費力地自己撰寫證明，這是一項不吸引人且耗時的任務。然而，在 2023 年，我有幸將這項任務委託給 ChatGPT。

All that was required from my end was to copy and paste the participant’s name and presentation title from our program and a brief description of the conference from our website. Everything else was handled by ChatGPT. Within a mere ten minutes, the certificate was generated, refined, and sent out-a staggering increase in efficiency of $83\mathrm{\%}$$83\mathrm{\%}$83%83 \%. One of the first studies of productivity gains from ChatGPT shows that, for mid-level professional writing tasks, the time spent decreases by $40\mathrm{\%}$$40\mathrm{\%}$40%40 \% while quality rises by $18\mathrm{\%}{}^1$$18\mathrm{\%}{}^1$18%^(1)18 \%{ }^{1} Those are very impressive numbers.
我所需做的只是從我們的計劃中複製並粘貼參與者的姓名和演講標題，以及從我們的網站上簡要描述會議。其他一切都由 ChatGPT 處理。在短短十分鐘內，證書就生成、完善並發送出去，效率驚人地提高了 $83\mathrm{\%}$$83\mathrm{\%}$83%83 \% 。關於 ChatGPT 生產力增長的第一批研究顯示，對於中層專業寫作任務，所花費的時間減少了 $40\mathrm{\%}$$40\mathrm{\%}$40%40 \% ，而質量則提高了 $18\mathrm{\%}{}^1$$18\mathrm{\%}{}^1$18%^(1)18 \%{ }^{1} 。這些數字非常令人印象深刻。

If you find it easy to dismiss this degree of efficiency gains as unworthy of exploration, then it is best to close this book now. There is little I could offer to sway your conviction. However, if you find such a prospect intriguing, I invite you to read on. This book will explore the transformative potential of AI in writing-related tasks for academic contexts. It will discuss how to navigate its challenges, and present a guide on how to integrate this technology responsibly and equitably into higher education.
如果你覺得輕易地忽視這種效率提升的程度是不值得探索的，那麼最好現在就關閉這本書。我無法提供什麼來改變你的信念。然而，如果你覺得這樣的前景引人入勝，我邀請你繼續閱讀。本書將探討人工智慧在學術環境中與寫作相關任務的變革潛力。它將討論如何應對其挑戰，並提供一個如何負責任和公平地將這項技術整合進高等教育的指南。

The advent of the AI era introduces “wraiting,” which is simply writing with the help of AI. It can also be called AI-assisted writing. If harnessed effectively, it can foster a synergy between human creativity and the unbridled computational power of AI. This novel practice presents a fascinating set of opportunities and challenges that are as intriguing as they are significant, compelling us to delve deep into the enthralling realm of wraiting.
AI 時代的來臨引入了“寫作”，這簡單地是指在 AI 的幫助下進行寫作。它也可以被稱為 AI 輔助寫作。如果有效利用，它可以促進人類創造力與 AI 無限制計算能力之間的協同作用。這一新穎的做法呈現出一系列既引人入勝又重要的機會和挑戰，迫使我們深入探索迷人的寫作領域。

This book postulates that wraiting is an advanced skill. Now, you might be thinking: why is something meant to simplify writing being presented as a
這本書假設寫作是一項高級技能。現在，你可能在想：為什麼一個旨在簡化寫作的東西會被呈現為一個

complex skill? Well, all tools designed to make our lives easier require understanding and skills to use effectively. Different people are adept at using different tools; some might be proficient with Adobe Photoshop, while others might be skilled with a race car or a violin. This great diversity of skills owes its existence to the need for specialized knowledge in every line of work.
複雜的技能？好吧，所有旨在讓我們生活更輕鬆的工具都需要理解和技能來有效使用。不同的人擅長使用不同的工具；有些人可能精通 Adobe Photoshop，而另一些人則可能擅長賽車或小提琴。這種技能的巨大多樣性源於每個行業對專業知識的需求。

The same pattern applies to this new tool, the AI-powered chatbot. It is undoubtedly easy to use. Simply go online, set up an account, and start typing or pasting your questions or requests. That is the beginning and end of the user manual: type or paste a question or request and press “Enter.” Compared to other information-technology innovations, this one has a remarkably low threshold for the initial use. However, using it to produce high-quality texts while also saving significant time is not so straightforward. While I do not consider myself a world champion in using AI-powered chatbots, my skills are better than those of most of my colleagues. As an educator and academic leader, I appreciate the trends, recognize the potential of this new tool, and understand its implications for higher education.
這種模式同樣適用於這個新工具，即人工智慧驅動的聊天機器人。它無疑易於使用。只需上網，設置一個帳戶，然後開始輸入或粘貼您的問題或請求。這就是用戶手冊的開始和結束：輸入或粘貼一個問題或請求，然後按“Enter”。與其他信息技術創新相比，這個工具的初始使用門檻顯著較低。然而，利用它來產生高質量的文本，同時節省大量時間，並不是那麼簡單。雖然我不認為自己是使用人工智慧驅動聊天機器人的世界冠軍，但我的技能比大多數同事要好。作為一名教育工作者和學術領導者，我欣賞這些趨勢，認識到這個新工具的潛力，並理解它對高等教育的影響。

However, another question emerges: is it worth investing time in mastering the more advanced uses of this new tool? Each person must answer this for themselves. I encourage pragmatism over prejudice. Try it first, and form an opinion later. Specifically, for educators and administrators, I see several compelling reasons to either invest time or at least consider the insights of this book:
然而，另一個問題出現了：花時間掌握這個新工具的更高級用法是否值得？每個人都必須自己回答這個問題。我鼓勵實用主義而非偏見。先試試，然後再形成意見。具體而言，對於教育工作者和管理者，我看到幾個令人信服的理由，無論是投資時間還是至少考慮這本書的見解：

Indeed, there are compelling reasons to invest the time, particularly if your professional role revolves around professional communications, scholarship, and teaching. Change is inevitable, and resistance to innovation can be costly to educators. Most importantly, it can be costly to our students. Resistance is driven by various factors, ranging from imminent retirement to an ideological stance. It is known from the earliest innovation studies that adoption of innovations is never uniform or simultaneous. ${}^2$${}^2$^(2){ }^{2} However, this particular innovation will be adopted quickly, because of the tremendous pressure from employers and students themselves. Its disruptive potential is very strong, and we in higher education may not have a choice in the matter.
確實，有充分的理由投入時間，特別是如果您的專業角色圍繞著專業溝通、學術研究和教學。變革是不可避免的，對創新的抵抗對教育工作者來說可能代價高昂。最重要的是，這對我們的學生來說也可能是代價高昂的。抵抗的原因各不相同，從即將退休到意識形態立場。從最早的創新研究中可以知道，創新的採用從來不是均勻或同時的。然而，這項特定的創新將會迅速被採用，因為來自雇主和學生本身的巨大壓力。它的顛覆潛力非常強大，而我們在高等教育中可能無法選擇這個問題。

Normally, I tend to hold off on learning new digital tools until I am convinced of their worth. There have been many tools that generated much hype, only to fade into the background. In the mid-2000s, a new technology, called “clickers” or audience response systems, created a wave of excitement in the world of education. They were touted as revolutionary tools that would make classes,
通常，我傾向於在確信新數位工具的價值之前不急於學習。曾經有許多工具引起了極大的熱議，卻最終淹沒在背景中。在 2000 年代中期，一種名為「點擊器」或觀眾反應系統的新技術在教育界引起了一陣興奮。它們被宣稱為革命性的工具，將使課堂變得更加生動。
especially large lectures, more interactive by allowing real-time, instant feedback. Schools invested heavily in these devices, and educational conferences buzzed with discussions about their potential. However, over time, the initial enthusiasm faded. Educators found that clickers did not themselves improve teaching or engagement. Instead, their effectiveness was tied to the quality of the pedagogical strategies employed. Additionally, logistical issues and the rise of multifunctional smartphones rendered these standalone devices redundant. By the late 2010 s, the hype around clickers had largely dissipated, teaching us that any new technology by itself is not a panacea, but merely a tool whose value is intrinsically tied to its application. This is just one example of the "tool fetishism ${}^{\prime \prime 3}$${}^{\prime \prime 3}$^(''3){ }^{\prime \prime 3} that is so typical for education technology.
尤其是大型講座，透過即時反饋使互動性更強。學校在這些設備上投入了大量資金，教育會議上充滿了對其潛力的討論。然而，隨著時間的推移，最初的熱情逐漸消退。教育工作者發現，點擊器本身並未改善教學或參與度。相反，它們的有效性與所採用的教學策略的質量息息相關。此外，後勤問題和多功能智能手機的興起使這些獨立設備變得多餘。到 2010 年代末，對點擊器的熱潮基本上消退，教會我們任何新技術本身並不是萬能的，而僅僅是一種工具，其價值本質上與其應用息息相關。這只是教育技術中典型的“工具迷戀”的一個例子。

With AI-powered chatbots, we are facing a phenomenon of a different order of magnitude. It represents a radical shift, larger than the transition from early search engines to Google, and larger than the introduction of word processing software. If you are old enough to remember these two, you can appreciate the scale of this change.
隨著人工智慧驅動的聊天機器人出現，我們面臨著一種不同層次的現象。這代表了一次根本性的轉變，規模超過了從早期搜尋引擎到谷歌的過渡，也超過了文字處理軟體的引入。如果你足夠年長，能夠記得這兩者，你就能夠體會這一變化的規模。
However, the decision to incorporate AI into our teaching practices should be a mindful, deliberate choice. To choose to do or not to do something, one needs to have a full picture of the option that is being embraced or rejected. In other words, if you know nothing about the use of $AI$$AI$AIA I in writing, your choice to reject is uninformed and arbitrary. That is another reason to learn about it.
然而，將人工智慧納入我們的教學實踐的決定應該是一個深思熟慮的選擇。要選擇做或不做某事，必須對所接受或拒絕的選項有全面的了解。換句話說，如果你對 $AI$$AI$AIA I 在寫作中的使用一無所知，那麼你拒絕的選擇就是不明智且隨意的。這是學習它的另一個理由。

This book is intended to help understand potential applications of AI-powered text-based tools in higher education. It is organized into five chapters: the first provides an overview of the fundamentals, while the subsequent chapters delve into the application of AI chatbots in three major domains: teaching, scholarship, and professional communications. The final chapter navigates ethical and philosophical considerations.
本書旨在幫助理解人工智慧驅動的文本工具在高等教育中的潛在應用。全書分為五個章節：第一章提供基本概念的概述，而隨後的章節則深入探討人工智慧聊天機器人在三個主要領域的應用：教學、學術和專業溝通。最後一章探討倫理和哲學考量。

Most of my experience lies with OpenAI’s ChatGPT model 4, the pioneering platform that unleashed the era of AI onto the mass market. I have also spent significant time with Claude 2, released in July of 2023. ${}^4$${}^4$^(4){ }^{4} Unlike its predecessor, which was not intended for mass use, Claude 2 can process much longer text inputs than ChatGPT. However, it appears less consistent in output quality and has a slower response time. As I will demonstrate throughout the book, some tasks are better handled by Claude, and others by ChatGPT. I utilize both daily, often in tandem, using the output of one to construct a prompt for the other.
我大部分的經驗來自於 OpenAI 的 ChatGPT 模型 4，這個開創性的平臺開啟了人工智慧進入大眾市場的時代。我也花了相當多的時間使用 Claude 2，這款於 2023 年 7 月發布的模型。與其前身不同，Claude 2 可以處理比 ChatGPT 更長的文本輸入。然而，它在輸出質量上似乎不那麼一致，且反應時間較慢。正如我在整本書中將要展示的，有些任務更適合由 Claude 處理，而另一些則由 ChatGPT 來完成。我每天都使用這兩者，經常是同時使用，利用一個的輸出來構建另一個的提示。

I have also explored Google’s Bard, another AI chatbot that operates similarly to ChatGPT. Bard’s constraints, particularly concerning the length of text it can generate, may limit its utility in certain scenarios, especially for longer-form content. Google has also integrated a limited AI capability into its Google Docs interface, allowing text within a document to be rewritten, or edited. While this is a promising and convenient interface that eliminates the need for multiple copying and pasting, the quality of its output is still behind that of the major competitors.
我也探索了谷歌的 Bard，這是一個與 ChatGPT 類似運作的 AI 聊天機器人。Bard 的限制，特別是在它能生成的文本長度方面，可能會在某些情況下限制其實用性，尤其是對於長篇內容。谷歌還在其 Google Docs 界面中集成了一個有限的 AI 功能，允許對文檔中的文本進行重寫或編輯。雖然這是一個有前景且方便的界面，消除了多次複製和粘貼的需要，但其輸出的質量仍然落後於主要競爭對手。

Microsoft’s Bing chatbot stands out with its real-time data access, distinguishing it from many counterparts. However, its utility for writing tasks seems less substantial compared to ChatGPT, Claude, or Bard. It might be better described as a supplement to a typical search engine rather than a full-fledged languagegenerating chatbot, with its strength lying more in data retrieval than creative writing or text generation. At present, ChatGPT is the most robust model in this field, although this status is by no means permanent. Claude’s unique ability to process extended inputs offers a promising direction for future development.
微軟的 Bing 聊天機器人因其即時數據訪問而脫穎而出，這使其與許多同類產品區別開來。然而，與 ChatGPT、Claude 或 Bard 相比，它在寫作任務中的實用性似乎不那麼顯著。它更像是一個典型搜索引擎的補充，而不是一個成熟的語言生成聊天機器人，其優勢更多體現在數據檢索而非創意寫作或文本生成。目前，ChatGPT 是這一領域中最強大的模型，儘管這一地位並非永久不變。Claude 處理擴展輸入的獨特能力為未來的發展提供了有希望的方向。

The bottom line in the rapidly evolving AI landscape is that ChatGPT’s influence has sparked promising innovations, but its long-term dominance is far from guaranteed. This competition should eventually lead to enhanced AI capabilities for consumers, but it is wise to temper expectations. It may take years of refinement and shifts in regulatory environments before any single contender can claim the title in the AI realm.
在快速發展的人工智慧領域，底線是 ChatGPT 的影響力激發了有前景的創新，但其長期主導地位遠未得到保證。這場競爭最終應該會為消費者帶來增強的人工智慧能力，但明智的做法是調整期望。可能需要數年的精煉和監管環境的變化，才能讓任何單一競爭者在人工智慧領域中聲稱擁有冠軍的頭銜。

This book serves as an initial investigation into the rapidly emerging field of AI-powered chatbots and their potential applications in higher education. Because these consumer-oriented technologies are still relatively new, there is a lack of extensive research and established best practices for their use. However, the potential for significant impact makes immediate study essential. To understand this phenomenon effectively, it is crucial to accumulate substantial experience using AI in higher education. This book aims to accelerate the accumulation of such practical experience so that it can be studied more rigorously.
本書作為對快速興起的人工智慧驅動聊天機器人及其在高等教育中潛在應用的初步研究。由於這些面向消費者的技術仍然相對新穎，因此缺乏廣泛的研究和既定的最佳實踐。然而，其潛在的重大影響使得立即進行研究變得至關重要。為了有效理解這一現象，積累在高等教育中使用人工智慧的豐富經驗至關重要。本書旨在加速這種實踐經驗的積累，以便能夠進行更嚴謹的研究。

My approach invites readers to participate in a shared experiment, exploring this uncharted domain through hands-on experiences rather than controlled studies. Rather than a trove of proven solutions, this book serves as an initial “treasure map,” pinpointing promising areas for exploration. It compiles hints, insights, and findings from my numerous micro-experiments in daily interactions with chatbots.
我的方法邀請讀者參與一個共同的實驗，通過實際體驗而非受控研究來探索這個未知的領域。這本書不是一個證明解決方案的寶庫，而是一個初步的“寶藏地圖”，標示出有前景的探索區域。它匯編了我在與聊天機器人日常互動中進行的眾多微實驗的提示、見解和發現。

Though light on citations due to the emerging nature of the field, this discovery-based approach sets the stage for future research by providing essential practical understanding. As we collectively uncover the benefits and effective applications of AI technology through shared experiences, we can move toward a future where AI chatbots enhance the field of higher education.
儘管由於該領域的發展性質而引用文獻較少，但這種基於發現的方法為未來的研究奠定了基礎，提供了必要的實踐理解。隨著我們共同通過共享經驗揭示人工智慧技術的好處和有效應用，我們可以朝著一個未來邁進，在這個未來中，人工智慧聊天機器人將提升高等教育領域。

In this section, I will present the concept of the “rich prompt”-the key ingredient in our recipe for successful “wraiting.”
在本節中，我將介紹“豐富提示”的概念——我們成功“寫作”配方中的關鍵成分。

One of the primary benefits of AI is that it does not ask us to master a distinct language to communicate with it. This has historically been, and continues to be, a steep barrier to using computers. While we have significantly moved away from the era of punch cards, interacting with a piece of software still requires us to learn its unique menus, interfaces, and conventions.
人工智慧的主要好處之一是它不要求我們掌握一種獨特的語言來與之交流。這在歷史上一直是，並且仍然是，使用電腦的一個重大障礙。儘管我們已經大幅擺脫了打孔卡的時代，但與一個軟體互動仍然需要我們學習其獨特的菜單、介面和慣例。

The AI-powered chatbot is designed to understand natural language, even if it is not expressed perfectly clearly. You can pose a direct question to it or ask it to generate a certain piece of text based on what you provide. However, there are better or worse questions or requests, and they will produce very different responses.
這個由人工智慧驅動的聊天機器人旨在理解自然語言，即使表達得不夠清晰。你可以直接向它提問，或要求它根據你提供的內容生成特定的文本。然而，問題或請求的好壞會影響其產生的回應，並且會產生非常不同的結果。

Let us commence with an essential reality check: AI does not possess the capability to “think” in the same way that humans do. It is best perceived as an amplifier for our thoughts, rather than an independent generator of ideas. It also can find and present ideas already expressed by others. The power of AI lies not in autonomous creation, but rather in the realms of effective communication and enhancement of ideas presented to it.
讓我們開始一個基本的現實檢查：人工智慧並不具備像人類一樣的“思考”能力。它最好被視為我們思想的放大器，而不是獨立的創意生成者。它也可以找到並呈現他人已經表達的想法。人工智慧的力量不在於自主創造，而是在於有效的溝通和增強呈現給它的想法。

When using AI, especially in contexts such as language generation, the quality of the output is intrinsically tied to the quality of the input. In other words, if you provide it with what I call a “lazy prompt,” you are likely to receive a mediocre or non-specific reply-this is often summed up as “garbage in-garbage out.” Sometimes the garbage looks fine at first glance. But read it more carefully, and you will see it is still just neatly packaged garbage.
使用人工智慧時，尤其是在語言生成等情境中，輸出的質量與輸入的質量本質上是息息相關的。換句話說，如果你提供一個我稱之為「懶惰提示」的內容，你很可能會收到一個平庸或不具體的回覆——這通常被總結為「垃圾進，垃圾出」。有時候，這些垃圾乍看之下似乎還不錯。但仔細閱讀後，你會發現它仍然只是包裝得很整齊的垃圾。

What I refer to as a “lazy prompt” is the act of pasting a task into a chatbot without any modification or additional guidance. Essentially, this is when you are not adding any personal input to the task. For example, a student might directly copy an essay assignment from a syllabus and paste it into a chatbot like ChatGPT, expecting a high-quality output. In the vast majority of situations, this kind of lazy prompting yields unsatisfactory results.
我所稱的「懶惰提示」是指將任務直接粘貼到聊天機器人中，而不進行任何修改或額外指導的行為。基本上，這是指你沒有對任務添加任何個人輸入。例如，一名學生可能會直接從課程大綱中複製一篇論文作業，然後將其粘貼到像 ChatGPT 這樣的聊天機器人中，期望得到高質量的輸出。在絕大多數情況下，這種懶惰提示會產生不令人滿意的結果。

However, there are rare instances where a lazy prompt might be justifiable. For example, if you are dealing with a bureaucratic task that you deem entirely meaningless, a lazy prompt might suffice for that specific context. But in most other scenarios you will need to invest your own thoughts or perspectives into the prompt to produce content that is actually valuable or meaningful. The key takeaway is that quality output usually requires quality input; you have to give a little to get something worthwhile in return.
然而，偶爾也有懶惰提示是可以被辯解的情況。例如，如果你正在處理一項你認為完全沒有意義的官僚任務，那麼在特定的情境下，懶惰提示可能就足夠了。但在大多數其他情況下，你需要投入自己的想法或觀點到提示中，以產生實際有價值或有意義的內容。關鍵在於，質量的輸出通常需要質量的輸入；你必須付出一些才能獲得值得的回報。

Conversely, when presented with a rich, well-crafted prompt, AI has the potential to generate something extraordinary, or at least deeply satisfying. It can weave together vast amounts of information, providing insights or compositions that might have taken much longer for a human to construct. Understanding the strengths and limitations of AI as a powerful tool, rather than an independent thinker, can enable us to utilize it more effectively in a multitude of applications.
相反，當面對一個豐富且精心設計的提示時，人工智慧有潛力產生一些非凡的東西，或至少是令人深感滿意的結果。它可以將大量信息編織在一起，提供洞見或作品，這些可能需要人類花費更長的時間來構建。理解人工智慧作為一種強大工具的優勢和局限性，而不是獨立思考者，可以使我們在多種應用中更有效地利用它。

Crafting rich prompts is an art, one that varies with context. Yet, despite their diversity, successful prompts share common elements. Let us delve into what constitutes a “rich” prompt, and how you can master the art of “wraiting.”
撰寫豐富提示是一門藝術，隨著上下文而變化。然而，儘管它們各不相同，成功的提示卻有共同的元素。讓我們深入探討什麼構成“豐富”的提示，以及你如何掌握“撰寫”的藝術。

The art of the rich prompt can be boiled down to these five points: give it an idea, give it a job, give it some food, give it a genre, and give it another push. Note that not every prompt has to meet all these criteria. In fact, you will see that a skillful use of conversations can reduce the prompts to a minimum. “Rich”
富有的提示藝術可以歸納為這五個要點：給它一個想法，給它一個任務，給它一些資料，給它一個類型，並再給它一個推動。請注意，並非每個提示都必須符合所有這些標準。事實上，您會發現熟練的對話使用可以將提示減少到最小。 “豐富”
does not mean “long.” It indicates quality, not quantity. Yet every good conversation with an AI-powered chatbot will start with a rich prompt.
並不意味著“長”。它表示質量，而不是數量。然而，與人工智慧驅動的聊天機器人的每一次良好對話都將以豐富的提示開始。

A rich prompt begins with a novel, unique idea that challenges AI to explore new avenues of thought. The more original and thought-provoking the idea, the more innovative and insightful the AI’s response is likely to be.
一個豐富的提示始於一個新穎、獨特的想法，挑戰人工智慧探索新的思維途徑。這個想法越原創、越引人深思，人工智慧的回應就越有創新性和洞察力。

AI models demonstrate proficiency in predicting likely sequences of text, adeptly encapsulating, and conveying widely accepted ideas. However, they consistently fall short of innovative thinking-an area traditionally within the purview of human intellect. Consequently, the first fundamental principle when using an AI chatbot is to recognize that the task of ideation-the birth of fresh concepts and content-remains squarely in the human domain, even though your own creativity may be stimulated by something AI generates.
AI 模型展示了預測文本可能序列的能力，熟練地概括和傳達廣泛接受的觀念。然而，它們在創新思維方面始終有所不足——這一領域傳統上屬於人類智慧的範疇。因此，使用 AI 聊天機器人的第一個基本原則是要認識到，創意的產生——新概念和內容的誕生——仍然完全在於人類的領域，即使你的創造力可能會受到 AI 生成內容的啟發。

In fact, if you encounter an output that presents something previously unknown to you, it is quite likely that it is reflecting existing knowledge of which you were simply unaware. It is true that AI can sometimes be nudged to inject a degree of randomness into its output, and, indeed, randomness can contribute an element of novelty. The degree to which novelty is attributed to randomness is a complex issue, which we will not delve into here. But it is important to remember that a well-presented point is not the same thing as an original idea, even though people routinely confuse the two.
事實上，如果你遇到一個輸出結果呈現了你之前未知的內容，那很可能是它反映了你只是未曾意識到的現有知識。確實，人工智慧有時可以被推動在其輸出中注入一定程度的隨機性，而隨機性確實可以為其增添一種新穎性。將新穎性歸因於隨機性的程度是一個複雜的問題，我們在這裡不會深入探討。但重要的是要記住，一個表達良好的觀點並不等同於一個原創的想法，儘管人們經常會混淆這兩者。

The tendency to conflate thinking and expression can potentially skew our understanding of these distinct processes. Thinking frequently occurs independently of linguistic expression. Some individuals might excel at articulating thoughts but struggle when it comes to creating original ideas. The rise of AIassisted writing can expose these weaknesses more prominently. One potential impact of the widespread use of chatbots could be that we will eventually sharpen our ability to differentiate between eloquent writing and original thought. Recognizing this distinction is a vital skill in both using AI tools effectively and appraising the intellectual contributions of others accurately.
將思考與表達混為一談的傾向可能會扭曲我們對這些不同過程的理解。思考經常獨立於語言表達進行。有些人可能在表達思想方面表現出色，但在創造原創想法時卻掙扎。AI 輔助寫作的興起可能會更明顯地暴露這些弱點。聊天機器人廣泛使用的一個潛在影響可能是，我們最終會提高區分優雅寫作與原創思想的能力。認識這一區別是有效使用 AI 工具和準確評估他人智力貢獻的重要技能。

Let us be real with ourselves: not all writing calls for originality. When it comes to the art of crafting prompts, context is king. Think back to that certificate of participation example from our introduction; it did not require any earthshattering ideas. In fact, much of our professional or academic communications does not demand a spark of creative thinking. Instead, it calls for alignment with established norms and conventions. Think of memos or reports-they are not typically vessels for groundbreaking concepts. In situations where compliance is paramount, being too original can, strangely enough, become a hurdle. But that is perfectly alright; it is simply the nature of those particular writing tasks.
讓我們對自己誠實：並非所有的寫作都需要創新。在撰寫提示的藝術中，背景是關鍵。回想一下我們介紹中的參與證書範例；它並不需要任何驚天動地的想法。事實上，我們的專業或學術交流中，很多時候並不需要創造性的思維火花。相反，它需要與既定的規範和慣例保持一致。想想備忘錄或報告——它們通常不是突破性概念的載體。在合規性至關重要的情況下，過於原創反而可能成為障礙。但這完全沒問題；這只是那些特定寫作任務的本質。

Regardless of the genre, whether it is a memo, an email, a report, or a piece of creative writing, it is essential to have a tight grip on the main idea or message
無論是什麼類型，無論是備忘錄、電子郵件、報告還是創意寫作，掌握主要思想或信息都是至關重要的
you are looking to convey. Understanding your intent will help you craft powerful prompts and, in turn, extract the most value from your AI tool. So, even if your writing does not call for original thought, clarity of purpose remains critical. That is the secret sauce to making AI work for you!
您希望傳達的內容。了解您的意圖將幫助您撰寫強有力的提示，進而從您的 AI 工具中提取最大的價值。因此，即使您的寫作不需要原創思考，明確的目的仍然至關重要。這就是讓 AI 為您服務的秘訣！

Setting up the “job” for an AI chatbot is a crucial step in achieving the desired results. Simply put, it is not just about giving the AI an idea; it is also about defining the work it has to do with that idea. You need to provide it with clear instructions and set it on a specific course.
為 AI 聊天機器人設置“工作”是實現期望結果的關鍵步驟。簡而言之，這不僅僅是給 AI 一個想法；還涉及到明確定義它需要用該想法完成的工作。您需要提供清晰的指示並為其設定具體的方向。

If you are drafting a scholarly article and need the AI to help improve your writing, a task could be “correct grammatical errors.” Here, you are clearly directing the AI to focus on grammar, to review your writing for errors, and to correct them. This task ensures that the AI does not stray into irrelevant territory, such as critiquing your argument or rearranging your logic.
如果您正在撰寫學術文章並需要人工智慧幫助改善您的寫作，任務可以是「糾正語法錯誤」。在這裡，您明確指示人工智慧專注於語法，檢查您的寫作是否有錯誤並進行修正。這個任務確保人工智慧不會偏離主題，例如批評您的論點或重新安排您的邏輯。

Alternatively, you might want the AI to enrich your text with examples and metaphors. In this case, the task could be “expand with more examples and metaphors.” This prompt guides the AI to not only stick to your idea but also support it with additional examples and metaphors, making your argument more vivid and engaging.
或者，您可能希望人工智慧用例子和隱喻來豐富您的文本。在這種情況下，任務可以是「用更多例子和隱喻擴展」。這個提示引導人工智慧不僅堅持您的想法，還用額外的例子和隱喻來支持它，使您的論點更加生動和引人入勝。

Finally, suppose you are involved in a debate and need the AI to help you think of counterarguments. The task could be “provide counterarguments to . . .” This request assigns the AI the job of thinking critically about your position and formulating plausible counterarguments.
最後，假設你參與了一場辯論，需要人工智慧幫助你思考反駁論點。這個任務可以是「提供對……的反駁論點」。這個請求將人工智慧的工作分配為批判性地思考你的立場並制定合理的反駁論點。

In each of these examples, the task you set significantly impacts the output from the AI. Active verbs such as “correct,” “expand,” and “provide” clarify the job for the AI, leading to a more focused and useful response. They constitute a necessary part of a rich prompt, leading to a more specific, appropriate, and helpful output from the AI.
在這些例子中，您設定的任務對 AI 的輸出有著重要影響。像“修正”、“擴展”和“提供”等主動動詞明確了 AI 的工作，從而導致更具針對性和實用性的回應。它們構成了豐富提示的必要部分，促使 AI 產生更具體、恰當和有幫助的輸出。

Starting with short paragraphs and specific ideas when writing with AI can be an effective strategy for generating high-quality content. To make the most of this approach, it can be helpful to use a range of productive prompts that encourage deeper thinking and more detailed analysis.
從短段落和具體想法開始，使用人工智慧寫作可以是一種有效的策略，以產生高品質的內容。為了充分利用這種方法，使用一系列能夠促進更深入思考和更詳細分析的有效提示會很有幫助。

Try using active verbs to describe the job you want AI to do for you, for example:
嘗試使用主動動詞來描述您希望人工智慧為您做的工作，例如：

main ideas. This can help you to build a more persuasive case and ensure that your content is backed up by solid research and analysis.
主要觀點。這可以幫助你建立更具說服力的論點，並確保你的內容有扎實的研究和分析作為支持。
3. Support by citing specific research: To further strengthen your arguments, you can ask the AI to provide specific research or studies that support your claims. This can help you to add credibility to your content and demonstrate that you have done your homework.
3. 透過引用具體研究來支持：為了進一步加強你的論點，你可以請求人工智慧提供支持你主張的具體研究或研究報告。這可以幫助你為你的內容增添可信度，並展示你已經做好了功課。
4. Illustrate with additional examples: Another useful prompt is to ask the AI to provide additional examples or case studies that illustrate your main points. This can help you to make your content more engaging and relatable to your readers.
4. 舉例說明：另一個有用的提示是請 AI 提供額外的例子或案例研究，以說明你的主要觀點。這可以幫助你使內容更具吸引力，並讓讀者更容易產生共鳴。
5. Provide analogies to elaborate: Analogies can be a powerful tool for explaining complex ideas in a more relatable way. You can ask the AI to provide analogies or metaphors that help to clarify your main points and make them more accessible to your audience.
5. 提供類比以進一步闡述：類比可以是解釋複雜概念的一個強大工具，能以更易於理解的方式呈現。您可以請求人工智慧提供類比或隱喻，以幫助澄清您的主要觀點，並使其對您的觀眾更具可接近性。
6. Present a counterargument to this: To ensure that your content is wellrounded and comprehensive, it can be helpful to ask the AI to provide counterarguments to your main points. This can help you to identify potential objections and weaknesses in your arguments and address them proactively in your content.
6. 提出反駁意見：為了確保您的內容全面且完整，請求人工智慧提供您主要觀點的反駁意見是很有幫助的。這可以幫助您識別潛在的反對意見和論點中的弱點，並在您的內容中主動解決這些問題。
7. Clarify/elucidate: If you find that your text is unclear or confusing, you can ask the AI to clarify specific points or ideas. For example, you could ask the AI to rephrase a sentence or to explain a concept in simpler terms.
7. 澄清/闡明：如果您發現您的文本不清楚或令人困惑，您可以請求人工智慧澄清特定的要點或想法。例如，您可以請求人工智慧重新表述一句話或用更簡單的術語解釋一個概念。
8. Summarize: To ensure that your content is easy to understand and digest, you can ask the AI to summarize specific sections or ideas. This can help you to condense complex information into a more concise and readable format.
8. 總結：為了確保您的內容易於理解和消化，您可以請求人工智慧總結特定的部分或想法。這可以幫助您將複雜的信息濃縮成更簡潔和易讀的格式。
9. Evaluate or critique: If you are unsure about the quality or effectiveness of your writing, you can ask the AI to evaluate specific sections or the overall piece. The AI could provide feedback on the tone, style, or clarity of your writing, or suggest ways to improve your arguments or presentation.
9. 評估或批評：如果您對自己的寫作質量或效果不確定，可以請人工智慧評估特定部分或整體作品。人工智慧可以提供有關您寫作的語氣、風格或清晰度的反饋，或建議改進您的論點或表達方式的方法。
10. Generate: Another useful prompt is to ask the AI to generate new ideas or content related to your topic. This can help you to explore new angles or perspectives on your subject matter and come up with fresh insights or approaches.
10. 生成：另一個有用的提示是請求人工智慧生成與您的主題相關的新想法或內容。這可以幫助您探索主題的不同角度或視角，並提出新的見解或方法。
11. Compare and contrast: If you are exploring multiple ideas or approaches, you can ask the AI to compare and contrast them. This can help you to identify the strengths and weaknesses of different arguments or strategies and determine which one is most effective.
11. 比較與對比：如果您正在探索多個想法或方法，您可以請 AI 進行比較與對比。這可以幫助您識別不同論點或策略的優缺點，並確定哪一個最有效。
12. Predict: If you are writing about a topic that involves future trends or developments, you can ask the AI to make predictions or projections. The AI could draw on data and trends to make informed guesses about what might happen in the future.
12. 預測：如果您正在撰寫有關未來趨勢或發展的主題，您可以請求人工智慧進行預測或推測。人工智慧可以根據數據和趨勢做出有根據的猜測，預測未來可能發生的事情。

The best help with AI is provided by AI itself. For example, it is always a good idea to use specific verbs while assigning a task. I was not sure how it interprets
最好的人工智慧幫助來自於人工智慧本身。例如，在分配任務時，使用具體的動詞總是一個好主意。我不確定它是如何解釋的。
the difference between “elucidate” and “expand.” Instead of experimenting with actual text, it is just easier to ask it. And here is the response:
“闡明”和“擴展”之間的區別。與其實際實驗文本，不如直接詢問。以下是回應：

When you use the command “elucidate,” I interpret it as a request to make something clearer, more understandable, or to provide further explanation on a particular point. The aim is to clarify the original meaning without necessarily adding much more content. It often involves explaining complex terms, providing additional context, or rephrasing for improved clarity.
當你使用“闡明”這個命令時，我將其解釋為要求使某事更清晰、更易理解，或對特定點提供進一步的解釋。目的是澄清原始意義，而不一定要添加更多內容。這通常涉及解釋複雜的術語、提供額外的背景信息或重新措辭以提高清晰度。

On the other hand, “expand” implies that you want me to add more information, detail, or breadth to the original content. It involves going beyond the initial statement, providing additional details, examples, or related information that could make the point more comprehensive or thorough.
另一方面，“擴展”意味著你希望我在原始內容中添加更多的信息、細節或廣度。這涉及超越最初的陳述，提供額外的細節、例子或相關信息，以使論點更加全面或徹底。

Note that ChatGPT and other chatbots understand repetition of prompts. If you want the same operation with the next fragment of test, just say “do the same with this . . . .” After a few repetitions, just paste the next segment without any prompt whatsoever.
請注意，ChatGPT 和其他聊天機器人能理解提示的重複。如果您想對下一段測試進行相同的操作，只需說「對這個做同樣的 . . . .」經過幾次重複後，只需直接粘貼下一段，而無需任何提示。

The practice of “wraiting” thrives on details, and, as such, it is necessary to feed the AI with ample information. Here, “Give it some food” is a metaphor that underscores the necessity of enriching your prompts.
“寫作”的實踐依賴於細節，因此，有必要為人工智慧提供充足的信息。在這裡，“給它一些食物”是一個隱喻，強調了豐富提示的必要性。

As you venture into the realm of AI-assisted writing, you may find that a sentence or two is simply not enough. Single sentences can be ambiguous, leading to outputs that may not align with your intent. On the other hand, rich prompts, akin to full paragraphs, paint a comprehensive picture of what you expect from the AI. They enable the AI to better understand the context, thereby reducing the chance of misinterpretation and increasing the likelihood of producing the desired response. If you are writing a blog with three points, list all three points, and beef them up with some detail. It is not always important which specific details, an example, or a hint will make the output better.
當你進入 AI 輔助寫作的領域時，你可能會發現一兩句話根本不夠。單一句子可能會含糊不清，導致產出的內容可能與你的意圖不符。另一方面，豐富的提示，類似於完整的段落，能夠全面描繪你對 AI 的期望。這使得 AI 能更好地理解上下文，從而減少誤解的機會，並增加產生所需回應的可能性。如果你正在撰寫一篇包含三個要點的博客，請列出所有三個要點，並用一些細節來充實它們。具體的細節、例子或提示並不總是重要，這些都能使產出更好。

In administrative communications, recycling existing texts as part of your prompt can often be a strategic approach to crafting effective communication. I will give more examples and guidance in Chapter Four. This does not mean copying and pasting text indiscriminately, but rather curating snippets that can guide the AI in the right direction. Whether it is a set of key points from a past report or snippets from well-received presentations, these texts can serve as a blueprint, helping the AI grasp the style, tone, and content you desire. Claude is especially adept at devouring large texts from which it can then harvest both contextual and factual information. Claude’s interface allows for file attachments, which I consider to be parts of the prompt. In ChatGPT, one needs to paste any “food” into the same prompt window. Keep in mind that ChatGPT will not process more than 2000 pasted words total in one prompt, so you may need to be more selective about how much it can swallow at one time. However, there is a way around it by uploading files.
在行政溝通中，將現有文本作為提示的一部分進行回收，通常是一種制定有效溝通的策略方法。我會在第四章提供更多示例和指導。這並不意味著不加區分地複製和粘貼文本，而是精心挑選可以引導人工智慧朝正確方向發展的片段。無論是來自過去報告的關鍵要點，還是來自受歡迎的演示文稿的片段，這些文本都可以作為藍圖，幫助人工智慧掌握您所期望的風格、語調和內容。Claude 特別擅長消化大量文本，從中提取上下文和事實信息。Claude 的界面允許文件附件，我認為這些是提示的一部分。在 ChatGPT 中，您需要將任何“食物”粘貼到同一提示窗口中。請記住，ChatGPT 在一個提示中不會處理超過 2000 個粘貼的單詞，因此您可能需要更有選擇性地決定一次可以提供多少內容。然而，通過上傳文件可以繞過這個限制。

To illustrate, let us consider a situation where you are crafting a grant proposal. You could incorporate extracts from a previous successful proposal, effectively training the AI on the desired structure, key points to include, perhaps borrowed from one of your publications, or from an email you write to your collaborator, and the tone that resonates with your audience. AI is omnivorous and will take everything into consideration. This way, you are not starting from scratch, but using established texts as scaffolding for new, yet consistent, content.
為了說明，讓我們考慮一個情境，您正在撰寫一份補助金提案。您可以納入之前成功提案的摘錄，有效地訓練人工智慧以達到所需的結構、關鍵要點，也許是借鑒自您的某篇出版物，或是您寫給合作者的電子郵件，以及與您的受眾共鳴的語氣。人工智慧是全能的，會考慮所有因素。這樣一來，您就不是從零開始，而是利用已建立的文本作為新內容的支架，且保持一致性。

Avoid the pitfall of excessive brevity and instead provide a comprehensive base for the AI to work from. By doing so, you will be setting the stage for the creation of content that is not only contextually accurate but also resonates with your target audience.
避免過度簡潔的陷阱，並提供一個全面的基礎讓人工智慧進行工作。這樣做將為創造不僅在語境上準確而且能引起目標受眾共鳴的內容奠定基礎。

In the quest to master the art of creating a rich prompt, the importance of specifying your audience, style, and genre cannot be overstated. This triumvirate forms the contextual backbone of your prompt, guiding the AI to tailor its output accordingly.
在掌握創建豐富提示藝術的過程中，明確指定你的受眾、風格和類型的重要性不容小覷。這三者構成了你提示的上下文支柱，引導人工智慧相應地調整其輸出。

Your audience could range from a group of corporate executives to middle school students, and their backgrounds, understanding, and interests significantly influence the kind of content that will resonate with them. A pointer to the AI on who the audience is can thus significantly enhance the relevance and effectiveness of the generated content. For instance, a report for a business audience may need to be filled with industry jargons, while an academic paper may demand a more scholarly language.
您的觀眾可能從一群企業高管到中學生不等，他們的背景、理解能力和興趣會顯著影響與他們產生共鳴的內容類型。因此，向人工智慧指明觀眾是誰，可以顯著提高生成內容的相關性和有效性。例如，針對商業觀眾的報告可能需要充滿行業術語，而學術論文則可能要求更學術化的語言。

The style of your content serves to refine the AI’s approach, determining the voice and tone of the written piece. For instance, a formal business report requires a professional and authoritative style, while a creative narrative may call for a more informal, expressive tone. By providing clear stylistic instructions to the AI, you ensure that the generated text aligns with your expectations, whether you desire a somber, humorous, casual, or scholarly tone.
您的內容風格有助於精煉人工智慧的處理方式，決定書面作品的聲音和語調。例如，正式的商業報告需要專業和權威的風格，而創意敘事則可能需要更非正式、表達豐富的語調。通過向人工智慧提供明確的風格指示，您可以確保生成的文本符合您的期望，無論您希望是嚴肅、幽默、隨意還是學術的語調。

In many cases, a brief description is enough to guide the AI. Typing commands like “write a blog” or “write a memo” will shift the output style to be more conversational or formal, respectively.
在許多情況下，簡短的描述足以指導人工智慧。輸入像是「寫一篇部落格」或「寫一份備忘錄」的指令，將使輸出風格分別轉變為更具對話性或正式性。

Furthermore, the genre acts as the framework for your content, defining its structure and conventions. Each type, whether it is a scientific research paper, business proposal, or fiction story, carries its own set of rules and expectations. Specifying the genre in your prompt helps the AI adhere to these conventions, ensuring the output is consistent with the norms of the chosen genre.
此外，類型作為您內容的框架，定義其結構和規範。每種類型，無論是科學研究論文、商業提案還是虛構故事，都有其自身的一套規則和期望。在您的提示中指定類型有助於人工智慧遵循這些規範，確保輸出與所選類型的標準一致。

The AI language models have a remarkable understanding of various text genres, thanks to their training on an extensive range of texts, including academic papers, emails, blog posts, and memos. Through this exposure, the models can discern the unique language patterns, structures, and conventions associated with each genre. As a result, AI can generate a text that precisely aligns with the
AI 語言模型對各種文本類型有著卓越的理解，這得益於它們在廣泛的文本範疇上進行的訓練，包括學術論文、電子郵件、博客文章和備忘錄。通過這種接觸，模型能夠辨別與每種類型相關的獨特語言模式、結構和慣例。因此，AI 能夠生成與該類型精確對應的文本。
specific genre you need. This ability is especially valuable in diverse writing scenarios, as it aids in shaping your writing to meet the conventions of the target genre.
特定類型的需求。這種能力在多樣化的寫作情境中尤其重要，因為它有助於塑造您的寫作，以符合目標類型的慣例。

Crafting a rich prompt is not a one-shot deal; it is often an iterative process. When you receive the AI’s response, take a moment to consider it, and then refine your prompt based on the outcome. This could involve further clarification of the task, adding more detail, adjusting the tone, or even pivoting the idea. This is the interaction between you and the AI -an ongoing exchange that brings the most effective content to life. You can ask the chatbot to redo, with adding new conditions, or you can revise the original prompt, and submit it again.
撰寫一個豐富的提示並不是一次性完成的；這通常是一個迭代的過程。當你收到人工智慧的回應時，花點時間考慮一下，然後根據結果來完善你的提示。這可能涉及進一步澄清任務、添加更多細節、調整語氣，甚至改變想法。這是你與人工智慧之間的互動——一個持續的交流，讓最有效的內容得以實現。你可以要求聊天機器人重新執行，並添加新的條件，或者你可以修訂原始提示，然後再次提交。

Improving your prompt is a skill, as delicate as any art form, which does not allow for shortcuts or simple recipes. It calls for practice, experimentation, and constant fine-tuning. As we step into the realm of AI -assisted writing, it is important to accept the unpredictability of AI-generated content. There is a black box between your input and the output, to which you have no access. The only way to improve quality is to tweak the input. The advanced skill lies not in memorizing a set of great prompts, but in the ability to improvise and refine prompts based on quick evaluation of the output.
改善你的提示是一種技巧，像任何藝術形式一樣精緻，並不允許捷徑或簡單的食譜。這需要練習、實驗和不斷的微調。當我們進入人工智慧輔助寫作的領域時，接受人工智慧生成內容的不可預測性是很重要的。在你的輸入和輸出之間有一個黑箱，你無法訪問。提高質量的唯一方法是調整輸入。高級技巧不在於記住一組優秀的提示，而在於根據對輸出的快速評估來即興創作和完善提示的能力。

When the AI’s output falls short of your expectations, revise the prompt. Is it clear? Does it have enough detail? Refining a less-than-perfect prompt could be the key to obtaining satisfactory results. However, if a well-crafted prompt still fails to yield the desired output, it might be an indication that the subject matter is an unfamiliar territory for the AI , or it got confused within the conversation.
當人工智慧的輸出未達到您的期望時，請修訂提示。它是否清晰？是否有足夠的細節？完善一個不夠完美的提示可能是獲得滿意結果的關鍵。然而，如果一個精心設計的提示仍然無法產生所需的輸出，這可能表明該主題對人工智慧來說是陌生的領域，或者它在對話中感到困惑。

In such cases, consider it as an opportunity to engage in original writing, rather than a frustration. By acknowledging AI’s limitations and seizing the chance to add our unique perspectives, we can leverage AI’s strengths to create work that is both innovative and engaging. As we continue to merge human creativity with AI , the evolving relationship between the writer and AI will open exciting new possibilities in writing.
在這種情況下，將其視為一個進行原創寫作的機會，而不是一種挫折。通過承認人工智慧的局限性並抓住機會加入我們獨特的觀點，我們可以利用人工智慧的優勢創造出既創新又引人入勝的作品。隨著我們不斷將人類創造力與人工智慧融合，作家與人工智慧之間不斷演變的關係將為寫作開啟令人興奮的新可能性。

The AI-enhanced writing process introduces a unique dynamic to this practice. The AI serves as a real-time feedback mechanism, offering insights and suggestions for refining prompts. As this interplay continues to evolve, we might see AI systems determining the best prompts for generating high-quality content, thereby refining their feedback mechanism to enhance the output. You can query, “How effective is my prompt? What additional information do you need to improve the output?” Often, the AI will provide a sensible recommendation.
AI 增強的寫作過程為這一實踐引入了獨特的動態。AI 作為一個實時反饋機制，提供見解和建議以完善提示。隨著這種互動的不斷演變，我們可能會看到 AI 系統決定生成高質量內容的最佳提示，從而完善其反饋機制以提升輸出。您可以詢問：“我的提示有多有效？您需要什麼額外信息來改善輸出？”通常，AI 會提供合理的建議。

I must highlight the fact that honing your skills in crafting prompts can be a significant boost for your growth not only as a “wraiter,” but also as a writer. As you continue refining your prompts, you will learn to differentiate more astutely between ideas and their expression. This practice compels you to be more precise and focused on your central message or concept, sharpening your skill in
我必須強調，磨練你撰寫提示的技能對於你作為一名「作家」以及作家的成長都是一個重要的助力。隨著你不斷完善你的提示，你將學會更敏銳地區分想法及其表達。這種練習迫使你更加精確和專注於你的核心信息或概念，提升你的技能。
conveying your thoughts with clarity. This is exactly why we in higher education should not only learn the skill of “wraiting” but also teach it to our students.
清晰地表達你的想法。這正是為什麼我們在高等教育中不僅應該學習“寫作”的技能，還應該教給我們的學生。

As we reach the conclusion of the discussion of rich prompts, I want to highlight the rapidly growing market of third-party add-ons or plug-ins that have been designed to assist users in their interactions with ChatGPT and other AI chatbots.
隨著我們討論豐富提示的結論，我想強調快速增長的第三方附加元件或插件市場，這些元件或插件旨在幫助用戶與 ChatGPT 及其他 AI 聊天機器人進行互動。

Without delving into specific product reviews, it is important to note that the quality and effectiveness of these tools can vary significantly. Some add-ons make grandiose promises that seem unrealistic, such as the ability to generate a full-length article from just a single keyword. Such promises may lead to disappointment as the outcomes often fall short of expectations.
在不深入探討具體產品評價的情況下，重要的是要注意這些工具的質量和效果可能會有很大差異。一些附加功能做出宏大的承諾，似乎不切實際，例如僅從一個關鍵字生成一篇完整的文章。這些承諾可能會導致失望，因為結果往往未能達到預期。

Others appear to add unnecessary complexity to otherwise straightforward tasks. For example, some plug-ins claim to facilitate the process of crafting a cover letter when users could simply ask the chatbot to create one by supplying their CV and job description. These intermediary tools sometimes operate by obscuring the actual prompts, which can hinder users from learning how to interact directly with the AI system.
其他人似乎為本來簡單的任務增加了不必要的複雜性。例如，一些插件聲稱能夠簡化撰寫求職信的過程，而用戶其實可以簡單地要求聊天機器人根據他們的履歷和職位描述來創建一封求職信。這些中介工具有時通過模糊實際的提示來運作，這可能會妨礙用戶學習如何直接與人工智慧系統互動。

There is a concern that relying too heavily on these intermediary tools may detract from the process of developing personal skills and forming a direct understanding of how to work with AI-powered tools. Critical thinking and original ideas remain paramount, both in content production and in the crafting of effective prompts. There is no substitute for these essential human abilities.
有一種擔憂是，過度依賴這些中介工具可能會削弱個人技能的發展過程，以及對如何使用人工智慧工具的直接理解。批判性思維和原創想法在內容創作和有效提示的制定中仍然至關重要。這些基本的人類能力是無法被替代的。

That said, it would be remiss not to acknowledge that some third-party plugins do provide valuable functionalities that overcome existing limitations in AI-powered chatbots. For example, certain add-ons enable ChatGPT to access the internet or larger files. Others assist in tasks like data analysis or accessing resources like Google Scholar. These tools can add value and enhance the user’s experience with the AI systems.
儘管如此，不承認某些第三方插件提供的有價值功能，克服了人工智慧聊天機器人的現有限制，將是失職的。例如，某些附加功能使 ChatGPT 能夠訪問互聯網或更大的文件。其他插件則協助進行數據分析或訪問像 Google Scholar 這樣的資源。這些工具可以增值並提升用戶與人工智慧系統的互動體驗。

The market for third-party add-ons is evolving at an incredible pace, making it challenging to predict what new functionalities may emerge soon. As this field continues to expand, it will be important for users to approach these tools with discernment, recognizing both the potential benefits and pitfalls. Emphasizing direct engagement, critical assessment, and ongoing learning will likely yield the most rewarding and effective experiences in working with AI chatbots.
第三方附加功能的市場正在以驚人的速度發展，這使得預測即將出現的新功能變得具有挑戰性。隨著這個領域的不斷擴展，使用者在使用這些工具時需要保持辨別力，認識到潛在的好處和陷阱。強調直接參與、批判性評估和持續學習，可能會在與人工智慧聊天機器人合作中帶來最有價值和有效的體驗。

OpenAI’s ChatGPT as well as Claude operates within the framework of “conversations” (also called “chats,” interchangeably). So, what exactly are these conversations and how can one make the best use of them?
OpenAI 的 ChatGPT 以及 Claude 都在「對話」（也稱為「聊天」，可互換使用）的框架內運作。那么，這些對話究竟是什麼，如何才能充分利用它們呢？

One intriguing characteristic of AI is its capacity for self-awareness, to a certain extent. This means that you can ask the AI directly about its functionalities
人工智慧的一個引人注目的特徵是其在某種程度上的自我意識。這意味著你可以直接詢問人工智慧有關其功能的問題。
when you are unsure. I frequently employ this method to enhance my understanding, just as I did to learn about “conversations.” Let me share what ChatGPT provided as a response when I queried about the concept of conversations:
當你不確定時。我經常使用這種方法來增強我的理解，就像我學習“對話”時所做的那樣。讓我分享當我詢問有關對話概念時，ChatGPT 提供的回應：

As an AI language model, my responses are generated based on the entire context of the conversation, not just the early parts. I use natural language processing and machine learning algorithms to analyze and understand the content of the conversation as it progresses, taking into account all the inputs and responses that have been exchanged so far. This allows me to generate responses that are contextually appropriate and relevant to the current stage of the conversation.
作為一個人工智慧語言模型，我的回應是根據整個對話的上下文生成的，而不僅僅是早期的部分。我使用自然語言處理和機器學習算法來分析和理解對話隨著進展而變化的內容，考慮到迄今為止所有的輸入和回應。這使我能夠生成在當前對話階段上下文上適當且相關的回應。

In other words, I do not specifically use inputs from the early part of a conversation to produce output for the later parts of the conversation. Instead, I use all the inputs and outputs from the entire conversation to generate responses that are appropriate for the current context. This approach allows me to generate responses that are coherent and relevant to the ongoing conversation, and to adapt to changes in the conversation as they occur. ${}^5$${}^5$^(5){ }^{5}
換句話說，我並不特別使用對話早期部分的輸入來產生對話後期部分的輸出。相反，我使用整個對話中的所有輸入和輸出來生成適合當前情境的回應。這種方法使我能夠生成與正在進行的對話一致且相關的回應，並隨著對話的變化而調整。

Similarly, if you attempt to reupload a document to Claude within the same chat, it will tell you that it can see the entire conversation and you do not need to upload it again.
類似地，如果您嘗試在同一聊天中重新上傳文件到 Claude，它會告訴您它可以看到整個對話，您不需要再次上傳。

In practical terms, it is advisable to keep separate projects or series of similar requests contained within their own unique conversations. This approach aids AI in contextual comprehension and enables it to carry over stylistic conventions set early in the conversation. For example, keep one conversation open for your emails and memos, another one for grant applications, and yet another for recommendation letters. The efficacy of AI often hinges on the length and depth of the conversation at hand. In essence, the more expansive the conversation, the more AI can grasp the context and nuances of your requests. Consider, for instance, a task as complex and creative as writing a book. It is highly beneficial to contain all related discussions, ideas, and drafts within one sustained conversation with the AI. If you find yourself content with the writing style and the output’s quality, maintaining the same conversation for a particular project can be useful, because it saves significant time on rewriting prompts.
在實際操作中，建議將不同的項目或類似請求的系列保留在各自獨特的對話中。這種方法有助於人工智慧理解上下文，並使其能夠延續在對話早期設定的風格慣例。例如，保持一個對話用於電子郵件和備忘錄，另一個用於資助申請，還有一個用於推薦信。人工智慧的效能往往取決於當前對話的長度和深度。從本質上講，對話越廣泛，人工智慧就越能掌握您請求的上下文和細微差別。舉例來說，寫一本書這樣複雜且富有創意的任務，將所有相關的討論、想法和草稿保留在與人工智慧的持續對話中是非常有益的。如果您對寫作風格和輸出質量感到滿意，對於特定項目保持相同的對話會很有幫助，因為這樣可以節省大量重寫提示的時間。

Essentially, think of conversations as context depositories. Once you have crafted your first rich prompt, there is no need to duplicate its depth with every subsequent prompt. You can simply instruct the AI to “apply the same style to this fragment,” and it will recall your previous prompts. The AI will preserve the style, the overall project’s context, and the terminology usage that you have previously established. It increases the value of your best prompts and makes them serve you longer.
本質上，將對話視為上下文的儲存庫。一旦您創建了第一個豐富的提示，就不需要在每個後續提示中重複其深度。您可以簡單地指示 AI “將相同的風格應用於這個片段”，它將回憶起您之前的提示。AI 將保留您之前建立的風格、整體項目的上下文和術語使用。這增加了您最佳提示的價值，並使它們能夠為您服務更長的時間。

During a conversation, you can leverage simpler instructions such as “elucidate,” “elaborate,” or “restate.” However, it is essential to remember that you should consistently provide the AI with portions of text to work with. In fact,
在對話中，您可以利用更簡單的指示，例如「闡明」、「詳細說明」或「重述」。然而，重要的是要記住，您應該始終提供給人工智慧一些文本供其處理。事實上，

after a few very similar prompts like “Elucidate,” the chatbot will assume that the next prompt is the same, so you can slide into a promptless request. This approach saves much time if you need to complete many repeated tasks. What I wrote in the section about the rich prompt applies to an entire conversation, not necessarily to every interaction with the conversation.
在幾個非常相似的提示如「闡明」之後，聊天機器人會假設下一個提示是相同的，因此你可以滑入無提示的請求。如果你需要完成許多重複的任務，這種方法可以節省很多時間。我在關於豐富提示的部分所寫的適用於整個對話，而不一定適用於與對話的每一次互動。

However, there might come an occasion when the AI appears to wrestle with a form of disarray. This typically surfaces during protracted conversations, particularly when you are incessantly tweaking sections of the same text. The AI may perceive your guidelines as too conflicting, leading to a state that we humanly interpret as “confusion.” From the AI’s perspective, however, it is more akin to a juggling act of incompatible instructions.
然而，可能會出現一種情況，當人工智慧似乎在某種混亂中掙扎。這通常在長時間的對話中出現，特別是當你不斷調整同一文本的各個部分時。人工智慧可能會將你的指導視為過於矛盾，導致一種我們人類解釋為「困惑」的狀態。然而，從人工智慧的角度來看，這更像是一種不相容指令的雜耍行為。

In the process of using AI , you continually revise the same text, each time issuing new, and potentially conflicting, instructions. The AI is programmed to interpret your prompts and react appropriately, all the while retaining knowledge of previous interactions within the conversation. However, when your instructions shift drastically or are contradictory, the AI may have difficulty in determining a coherent course of action.
在使用人工智慧的過程中，您不斷修訂相同的文本，每次發出新的、可能相互矛盾的指示。人工智慧被編程為解釋您的提示並做出適當反應，同時保留對對話中先前互動的知識。然而，當您的指示發生劇烈變化或相互矛盾時，人工智慧可能會難以確定一致的行動方針。

In these situations, the most prudent course of action may be to relinquish the current conversation and initiate a brand-new one. A fresh conversation functions as a clean slate, devoid of the contradictory elements that clouded the previous discourse. However, it is essential to keep in mind that, within a new conversation, you should reconstruct the elements of a rich prompt. Anytime a chatbot stops making sense, try a new conversation first. If it still does not make sense, the problem is with your prompts.
在這些情況下，最明智的行動可能是放棄當前的對話，並開始一個全新的對話。全新的對話就像一張乾淨的白紙，沒有之前討論中混淆的矛盾元素。然而，重要的是要記住，在新的對話中，您應該重建豐富提示的元素。每當聊天機器人開始變得不合邏輯時，首先嘗試一個新對話。如果它仍然沒有意義，那麼問題出在您的提示上。

While AI technology, at this point, is not capable of being tailored precisely to your individual preferences and cannot yet learn your individual writing style and voice, a continuous and expansive conversation is the nearest approximation to such personalization. The AI learns and adjusts based on the cumulative inputs received within the context of the conversation, which can yield results that are surprisingly attuned to your unique needs.
雖然目前的人工智慧技術尚無法精確地根據您的個人偏好進行調整，也尚未能學習您的個人寫作風格和聲音，但持續且廣泛的對話是最接近這種個性化的方式。人工智慧會根據在對話過程中收到的累積輸入進行學習和調整，這可能會產生出意外符合您獨特需求的結果。

However, for standalone inquiries or tasks, it might be beneficial to delete the conversation once it has served its purpose, so as not to clutter your workspace. This ensures that your interface remains clean and navigable, enhancing your efficiency and productivity.
然而，對於獨立的查詢或任務，刪除對話一旦完成其目的可能是有益的，以免使您的工作空間變得雜亂。這確保了您的介面保持整潔和易於導航，從而提高您的效率和生產力。

You will learn to keep some good conversations going for a long time, while others are not worth keeping around.
你將學會如何讓一些良好的對話持續很長時間，而另一些則不值得保留。

When defining the desired writing style, we face a dilemma. Without clear instructions, the output can be generic, clichéd, and lacking individuality. While this may suffice in certain contexts, such as generic business writing, where conformity to conventions is more important than personal expression, it falls
在定義所需的寫作風格時，我們面臨一個兩難的局面。若沒有明確的指示，產出的內容可能會顯得普通、陳腔濫調，且缺乏個性。雖然在某些情境下，例如普通的商業寫作中，遵循慣例比個人表達更為重要，但這樣的寫作風格卻顯得不足。
short in many other instances. Remember that the language model is trained to recall the most likely combination of words. Therefore, cliché and commonplace expressions are its trademark. On the other hand, if we provide the AI with excessive style directions, we risk creating writing that appears forced and ridiculous. It is a fine line to tread.
在許多其他情況下都很簡短。請記住，語言模型是訓練來回憶最可能的單詞組合。因此，陳詞濫調和普通表達是它的標誌。另一方面，如果我們給予人工智慧過多的風格指導，我們就有可能創造出看起來生硬和可笑的寫作。這是一條微妙的界線。

Looking ahead, the prospect of training an AI to mimic our personal writing style is undeniably enticing. However, the current limitations on input data size pose an obstacle.
展望未來，訓練一個人工智慧模仿我們個人寫作風格的前景無疑是令人嚮往的。然而，目前對輸入數據大小的限制則構成了一個障礙。

Replicating a personal writing style using AI involves several approaches, each with varying degrees of success. One approach is to use a sample of personal writing as a style guide, feeding it into the AI along with unrelated content, with the expectation that the AI will borrow stylistic elements while disregarding the actual content of the guide. This method can fall short as the AI might integrate actual content from the guide text, undermining the goal of contentindependent style replication.
使用人工智慧複製個人寫作風格涉及幾種方法，每種方法的成功程度各不相同。一種方法是使用個人寫作的樣本作為風格指南，將其與無關的內容一起輸入 AI，期望 AI 能夠借用風格元素，同時忽略指南的實際內容。然而，這種方法可能會失敗，因為 AI 可能會整合指南文本中的實際內容，從而削弱了獨立於內容的風格複製的目標。

Another strategy involves initially presenting the AI with a piece of personal writing and instructing it to both read and describe the writing style contained within. Using the AI-generated description of the style as a blueprint for subsequent pieces can be more effective. However, this approach might still struggle to capture the distinctive “voice” that characterizes individual writing. It may lack the depth and intricacies that arise from direct perception, leading to an output that pales in comparison to the original.
另一種策略是最初向人工智慧提供一段個人寫作，並指示它閱讀並描述其中的寫作風格。使用人工智慧生成的風格描述作為後續作品的藍圖可能會更有效。然而，這種方法仍然可能難以捕捉到個人寫作所特有的獨特“聲音”。它可能缺乏直接感知所帶來的深度和複雜性，導致產出的作品與原作相比顯得黯然失色。

ChatGPT has recently been employing the so-called “customized instructions,” allowing users to store a general description of their preferences for output style. This is a positive development and a step in the right direction. Hopefully, advancements in this direction will continue, leading to more refined and personalized interactions with the AI.
ChatGPT 最近開始使用所謂的「自訂指示」，允許用戶儲存對輸出風格的一般描述。這是一個積極的發展，也是朝著正確方向邁出的一步。希望在這方面的進展能夠持續，帶來與人工智慧更精緻和個性化的互動。

A more effective approach might involve identifying a body of text closely aligned with the desired stylistic preferences and instructing the AI to mirror that style. Directing the AI to emulate the style of essays from renowned publications or drawing inspiration from prolific writers can yield better results. However, even this method reveals the complexity of truly replicating an individual’s unique writing style, demonstrating that there are still limitations and challenges in the field of AI-generated content.
一種更有效的方法可能涉及識別一段與所需風格偏好密切相關的文本，並指示人工智慧模仿該風格。指導人工智慧模仿知名出版物的文章風格或從多產作家那裡汲取靈感可以產生更好的結果。然而，即使這種方法也顯示出真正複製個人獨特寫作風格的複雜性，顯示出在人工智慧生成內容的領域中仍然存在限制和挑戰。

It is crucial to exercise caution with this approach. If you instruct the AI to write like specific authors such as David Remnick or Ernest Hemingway, the output may resemble a parody, uncannily close but somehow missing the mark. Instead, it might be more advisable to select a broader corpus of works from several authors whose writing styles you admire. For example, consider giving an instruction like this: “Make it sound like a typical The New Yorker non-fiction essay.”
在這種方法上謹慎行事是至關重要的。如果你指示人工智慧模仿特定作者，如大衛·雷姆尼克或歐內斯特·海明威，產出的內容可能會像是一種模仿，雖然非常接近但卻有些偏離目標。相反，選擇幾位你欣賞的作者的更廣泛作品集可能更為明智。例如，可以考慮給出這樣的指示：“讓它聽起來像是一篇典型的《紐約客》非虛構文章。”

This strategy brings up an interesting ethical question about the emulation of someone else’s writing style. Is this not something we do during our education?
這個策略提出了一個有趣的倫理問題，即模仿他人的寫作風格。這不正是我們在教育過程中所做的事情嗎？

Do we not read books and learn how to write from others? Have not many great writers acknowledged their debt to others they learned from?
我們難道不閱讀書籍並向他人學習寫作嗎？許多偉大的作家難道沒有承認他們對學習對象的感激之情嗎？

By studying and incorporating the collective influences of multiple authors who share similarities with your preferred style, you can avoid the pitfalls of imitating a single author too closely. This method leverages the breadth of their work to develop distinctive AI-generated output that genuinely reflects your voice, striking a balance between drawing inspiration and maintaining authenticity.
通過研究並融入多位與您喜好風格相似的作者的集體影響，您可以避免過於緊密模仿單一作者的陷阱。這種方法利用他們作品的廣度來發展獨特的 AI 生成輸出，真正反映您的聲音，在汲取靈感和保持真實性之間取得平衡。

Another trick is to start each segment you wish to edit with a well-written paragraph in your preferred style. Include in your prompt an instruction like this: “Use the first paragraph as the guiding style sample.”
另一個技巧是以一段寫得好的段落開始您希望編輯的每個部分，並使用您喜歡的風格。在您的提示中包含這樣的指示：“將第一段作為指導風格範本。”

However, this approach is not without its challenges. Manual editing may be necessary to refine elements that fall short of an authentic personal voice. In fact, one of the advanced skills at which humans so far outperform the AI is the ability to tweak style, to create a unique voice.
然而，這種方法並非沒有挑戰。手動編輯可能是必要的，以完善那些未能達到真實個人聲音的元素。事實上，人類迄今為止在一項高級技能上遠超人工智慧，那就是調整風格、創造獨特聲音的能力。

There is a threshold beyond which crafting the piece personally becomes more efficient. The decision hinges on the significance you attribute to your individual voice for the specific writing endeavor. Perhaps voice is less important in an accreditation report, so you can ignore the style altogether ю
有一個門檻，超過這個門檻，親自創作這件作品會變得更有效率。這個決定取決於你對於特定寫作任務中個人聲音的重要性。也許在認證報告中，聲音的重要性較低，因此你可以完全忽略風格。

Although AI can save significant time on text generation, expect to spend a bit more time on editing. The overall time savings are still very impressive but do allocate enough time for re-reading and editing the final product. We will return to this important point in the section on authorship. The authorial responsibility is shifting from the act of text generation to the act of the release of the text to the world.
雖然人工智慧可以在文本生成上節省大量時間，但預期在編輯上會花費更多時間。整體的時間節省仍然非常令人印象深刻，但請確保為重新閱讀和編輯最終產品分配足夠的時間。我們將在關於作者身份的部分回到這個重要的點。作者的責任正從文本生成的行為轉移到將文本釋放到世界的行為上。

In this captivating exploration of personalizing AI writing, we balance efficiency and authenticity. As we move closer to a future where AI can faithfully reflect our individual writing styles, the inherent, irreplaceable value of the human touch in writing must be acknowledged. Despite AI’s remarkable advancements, the idiosyncratic beauty of human expression remains an art form of its own, illustrating that some nuances of personal style may forever lie beyond the reach of even the most advanced computers.
在這個引人入勝的探索中，我們平衡了個性化 AI 寫作的效率與真實性。隨著我們逐漸接近一個 AI 能夠忠實反映我們個人寫作風格的未來，必須承認人類在寫作中所具備的固有且無法取代的價值。儘管 AI 的驚人進步，人類表達的特有之美仍然是一種獨特的藝術形式，這表明某些個人風格的細微差別可能永遠超出即使是最先進的計算機的觸及範圍。

The experience of writing with AI, or what I like to call “wraiting,” is an adventure into uncharted territories, a type of experience that may initially feel unfamiliar and challenging. Ironically, the better a writer you are, the more confusing AI assistance may initially feel. Nevertheless, this undertaking is not as daunting as rocket science, and it is important to approach it with an open mind, curiosity, and a sense of adventure. This section is my attempt to prepare you psychologically for learning how to “wraite.”
與人工智慧寫作的經驗，或我喜歡稱之為「寫作」，是一場探索未知領域的冒險，這種經驗最初可能會感到陌生且具有挑戰性。諷刺的是，寫得越好，人工智慧的協助最初可能會讓人感到越困惑。然而，這項工作並不像火箭科學那麼艱巨，重要的是以開放的心態、好奇心和冒險精神來面對它。本節是我試圖在心理上為你準備學習如何「寫作」。

Once upon a time, writing was a solitary endeavor. But with AI’s involvement, it has transformed into a creative partnership, where you are not the only participant. Although some might perceive this change as dystopian, I view it as
從前，寫作是一項孤獨的工作。但隨著人工智慧的參與，它已轉變為一種創意夥伴關係，在這裡你不是唯一的參與者。雖然有些人可能將這種變化視為反烏托邦，但我卻將其視為
an era of “wraiting” - a harmonious blend of traditional writing and AI, forming an intriguing symbiotic relationship. Only very few people who have experience working with ghostwriters and speechwriters have something in their experience to prepare them for this shift.
一個“寫作”的時代——傳統寫作與人工智慧的和諧融合，形成了一種引人入勝的共生關係。只有極少數有與代筆者和演講稿撰寫者合作經驗的人，才能在他們的經歷中為這一轉變做好準備。
“Wraiting” requires one to set aside any anxieties about the content generated. One needs to overcome the traditional demarcation between “mine” and “not mine.” When text is crafted from your AI-guided prompts, even if it does not mirror your usual style or voice, it is vital to perceive it as yours. Or, rather, what is important and original there is yours, and whatever is not yours is common knowledge. Consider this as interacting with an alternate version of yourself, a digital reflection contributing to your creative process. This way, you can fully embrace the potential of writing and let this new form of expression enhance your creativity.
“寫作”需要人們放下對生成內容的任何焦慮。人們需要克服“我的”和“不是我的”之間的傳統界限。當文本是根據你的 AI 指導的提示創作時，即使它不反映你平常的風格或聲音，重要的是要將其視為你的作品。或者，更確切地說，那裡重要且原創的部分是你的，而不屬於你的部分則是常識。將這視為與你自己的一個替代版本互動，一個數位反映，為你的創作過程做出貢獻。這樣，你就可以充分擁抱寫作的潛力，讓這種新的表達形式增強你的創造力。

The first time you try “wraiting” with a chatbot, it can feel weird. When the chatbot quickly produces a well-written piece, you might think, “Wow! I have spent years learning how to write well, and this machine can do something similar in no time!” It is surprising and a little unsettling to see a machine do something you have worked hard to be good at.
第一次與聊天機器人嘗試「寫作」時，可能會感到奇怪。當聊天機器人迅速產出一篇寫得很好的文章時，你可能會想：「哇！我花了多年時間學習如何寫好，而這台機器卻能在短時間內做到類似的事情！」看到一台機器做你努力想要擅長的事情，令人驚訝且有些不安。

This unsettling feeling is not unique to our tech-infused era, nor is it exclusive to the white-collar demographic. Let us turn the clock back to the 18th and 19th centuries, during the heyday of the Industrial Revolution. Craftsmen, renowned for their skills and the impeccable quality of their handiwork, found their lives upended by the advent of mass production.
這種不安的感覺並非我們這個科技融入的時代所獨有，也不僅限於白領階層。讓我們回到 18 世紀和 19 世紀，工業革命的全盛時期。以技藝高超和手工藝品質量無可挑剔而聞名的工匠，因大規模生產的到來而生活被顛覆。

The birth of factories manufacturing goods at a fraction of the price of artisanal products posed an existential threat to these craftsmen. Consider the humble shoemaker who crafted footwear with painstaking precision, suddenly confronted by an avalanche of shoes churned out by mechanical marvels at prices his handcrafted shoes could not compete with. This shift exerted immense economic pressure, compelling craftsmen to either adapt their skills to the new industrial world order or seek alternate professions. We do not have to travel too far back; several occupations have shrunk or disappeared during our lifetimes: travel agents, video store clerks, photo film developers, map printers, phone book publishers, and bank tellers. How do you think all these people felt? Naturally, their initial reaction was one of disbelief and outright denial.
工廠以極低的價格生產商品的誕生對這些工匠構成了生存威脅。想想那位謙遜的鞋匠，他以極大的精確度製作鞋子，卻突然面對著一波波由機械奇蹟生產的鞋子，價格是他手工製作的鞋子無法競爭的。這一變化施加了巨大的經濟壓力，迫使工匠們要麼調整自己的技能以適應新的工業世界秩序，要麼尋求其他職業。我們不必回溯太久，幾個職業在我們的生命中已經縮減或消失：旅行社、錄影帶店店員、照片沖洗技術員、地圖印刷商、電話簿出版商和銀行出納員。你認為這些人感覺如何？自然，他們最初的反應是難以置信和完全否認。

Yet, as I navigated through my initial wave of discomfort with AI, I noticed its limitations and constraints. And I will confess that this revelation brought with it a sigh of relief. Recognizing AI’s weaknesses made me realize that I possess something that it does not. Thus, the initial and perhaps the most crucial step is to maintain a calm and collected demeanor in the face of AI. Staying composed enables further learning. It means avoiding panic and not dismissing AI as a trivial gimmick. Navigating between the extremes of fear and denial allows you to weather this period of transition and progress further into the intriguing world of “wraiting.”
然而，當我在最初對人工智慧的不適中摸索時，我注意到了它的局限性和約束。我必須承認，這一啟示帶來了一絲鬆口氣。認識到人工智慧的弱點讓我意識到我擁有一些它所不具備的東西。因此，最初也是最關鍵的一步是在面對人工智慧時保持冷靜和沉著。保持鎮定使進一步學習成為可能。這意味著避免恐慌，而不是將人工智慧視為微不足道的花招。在恐懼和否認的極端之間導航，讓你能夠度過這一過渡期，並進一步進入“寫作”的迷人世界。

The novelty of AI tools tends to fade as you become more familiar with their use. With about a month’s worth of daily practical use, these tools become less of
隨著你對這些工具的使用越來越熟悉，人工智慧工具的新穎感往往會減退。經過大約一個月的每日實際使用，這些工具變得不那麼令人驚訝。

a wonder and more of an everyday utility. You start understanding their capabilities and limitations more deeply, and the output they generate becomes increasingly predictable. The initial sense of magic gradually dissipates, reinforcing the fact that, at its core, an AI is ultimately just a cleverly designed instrument.
一種奇蹟，更是一種日常工具。你開始更深入地理解它們的能力和局限性，而它們產生的輸出變得越來越可預測。最初的魔力感逐漸消散，強化了這個事實：在其核心，人工智慧最終只是一種巧妙設計的工具。

As I ventured deeper into the world of “wraiting,” I began to acknowledge that crafting sentences, while a cornerstone of writing, could sometimes be a laborious task worthy of being outsourced. This realization comes easier with professional communications such as reports, applications, recommendations, and letters, where personal voice often takes a backseat to the conventional structure and format. As I entrusted these tasks to AI, my frustrations over maintaining voice and authorship receded, providing the cognitive schemata to differentiate between those aspects of writing that can be delegated to AI and those that demand human mind.
隨著我深入探索“寫作”的世界，我開始意識到，雖然構建句子是寫作的基石，但有時這可能是一項值得外包的繁重任務。這種認識在專業通信中更容易出現，例如報告、申請、推薦信和信件，在這些情況下，個人聲音往往退居於傳統結構和格式之後。當我將這些任務委託給人工智慧時，我對保持聲音和作者身份的挫折感減少了，這提供了認知框架，以區分那些可以委派給人工智慧的寫作方面和那些需要人類思維的部分。

Interestingly, using AI to handle some of the basic writing tasks did not make me care less about the quality; it made me focus on it more. By letting the AI do some of the initial work, I had more energy to look closely at the details of the text. I spent more time than before considering word choices, sentence structure, and style. This did not show a problem with the AI, but rather gave me something valuable: more time. The time I used to spend on the basic writing tasks could now be used for careful editing. In this way, using “wraiting” saved me time on the initial steps, allowing me to put more effort into improving the final product.
有趣的是，使用人工智慧處理一些基本的寫作任務並沒有讓我對質量漠不關心；反而讓我更加專注於質量。讓人工智慧完成一些初步工作後，我有更多精力仔細查看文本的細節。我花了比以前更多的時間考慮用詞、句子結構和風格。這並不是顯示人工智慧的問題，而是給了我一些寶貴的東西：更多的時間。我以前用來處理基本寫作任務的時間現在可以用來進行仔細的編輯。這樣一來，使用“寫作”讓我在初步步驟上節省了時間，使我能夠將更多精力投入到改善最終產品中。

Yet, it is important to note that my journey and the epiphanies along the way are not universal blueprints. They signify a personal choice, a distinctive trail I have elected to traverse. Not everyone may find the same delight or sense of creative emancipation through “wraiting.” But its potential is undeniable"wraiting" can redefine our relationship with the written word, economizing our efforts on foundational tasks and enabling us to ascend toward more advanced creative heights.
然而，重要的是要注意，我的旅程和沿途的頓悟並不是普遍的藍圖。它們代表了一個個人的選擇，是我選擇走過的獨特道路。並不是每個人都能通過“wraiting”找到相同的樂趣或創造性的解放。但其潛力是不可否認的，“wraiting”可以重新定義我們與書面文字的關係，節省我們在基礎任務上的努力，使我們能夠向更高級的創造性高度邁進。

AI, always polite and calm, can be likened to a personal assistant. The assistant may not be the brightest you ever had, but it is incredibly fast, is very well informed in all disciplines and fields, does not get annoyed or tired, and does not ask for a paycheck. Where humans get tired of repetitive tasks, AI only learns more. It is always ready to perform tasks that I would rather not do myself. The overall experience is more liberatory than confining.
人工智慧總是禮貌且冷靜，可以比擬為個人助理。這位助理可能不是你曾經擁有的最聰明的，但它非常快速，對所有學科和領域的知識非常豐富，不會感到煩惱或疲倦，也不會要求薪水。人類在重複性任務中感到疲倦，而人工智慧則只會學習得更多。它總是準備好執行我不想自己做的任務。整體體驗比起束縛來說更具解放性。

One critical aspect to be cognizant of is understanding the optimal text size for interactions with AI chatbots. Knowing the size limitations of text that these AI models can handle is crucial to obtaining high-quality outputs and ensuring efficient interaction. I hope this is a temporary condition, but it is also likely that scarcity of processing capacity will present an ongoing natural limit to expanding the text size capacity. The input and output sizes are two very different things.
一個需要注意的關鍵方面是了解與人工智慧聊天機器人互動的最佳文本大小。了解這些人工智慧模型能處理的文本大小限制對於獲得高質量的輸出和確保高效的互動至關重要。我希望這是一個暫時的情況，但處理能力的稀缺也可能會對擴大文本大小的能力形成持續的自然限制。輸入和輸出大小是兩個非常不同的概念。

If you have tried to feed in a text longer than ChatGPT’s comfort zone, which is approximately 2000 words, you would likely receive a response indicating that the text is too long. As users, we need to be mindful of these constraints to avoid input overload and to ensure that our AI companion works in its most effective range. Starting with October 2023, ChatGPT allows limited file uploads. Claude, however, will accept very large texts, in various file formats, up to 5 files, 10 Mb each. It is coy about disclosing the exact word count limitation, but it can take 50,000 words in plain text for a high-level analysis. It will not edit, or revise the entire text, but can provide a critique or write a review, an outline, or suggest improvements.
如果您嘗試輸入一段超過 ChatGPT 舒適範圍的文本，該範圍大約為 2000 字，您可能會收到一個回應，指出文本太長。作為用戶，我們需要注意這些限制，以避免輸入過載，並確保我們的 AI 夥伴在其最有效的範圍內運作。從 2023 年 10 月開始，ChatGPT 允許有限的文件上傳。然而，Claude 可以接受各種文件格式的非常大文本，最多可上傳 5 個文件，每個文件 10 Mb。它對於具體的字數限制保持含糊，但可以接受 50,000 字的純文本進行高級分析。它不會編輯或修訂整個文本，但可以提供批評或撰寫評論、大綱，或建議改進。

Generally, both ChatGPT and Claude tend to produce outputs of about 500600 words. With some prodding, Claude can generate a 1500 -word text, but it may be diminished in quality. You might ask, “Is this a hard limit?” The answer is no; it is not a rigid restriction. Instead, think of it as a rule of thumb or a guide that helps you better utilize ChatGPT or Claude. When we provide our AI tools with information within their optimal processing range, we are more likely to get outputs that are both accurate and useful.
一般來說，ChatGPT 和 Claude 的輸出通常約為 500 到 600 字。經過一些刺激，Claude 可以生成 1500 字的文本，但質量可能會下降。你可能會問：“這是一個硬性限制嗎？”答案是否定的；這不是一個嚴格的限制。相反，將其視為一條經驗法則或指導，幫助你更好地利用 ChatGPT 或 Claude。當我們向我們的 AI 工具提供在其最佳處理範圍內的信息時，我們更有可能獲得準確且有用的輸出。

Imagine you have a comprehensive report or a lengthy essay of about 6000 words that you want ChatGPT to process. To handle this efficiently, you would need to divide the document into smaller chunks because the optimal output lies somewhere in the ballpark of 600 words. This means you should think about breaking it down into approximately 10 chunks in terms of output. If you use the technique of “sloppy jotting” that I describe later in the book, you will have to produce inputs that will be OK after growing to about 500-600 words.
想像一下，您有一份約 6000 字的綜合報告或長篇論文，希望 ChatGPT 進行處理。為了有效處理，您需要將文件分成較小的部分，因為最佳輸出大約在 600 字左右。這意味著您應該考慮將其分解為大約 10 個部分。如果您使用我在書中稍後描述的“隨意記錄”技術，您將需要產生在增長到約 500-600 字後仍然可以接受的輸入。

At first glance, this might seem like an extra step. However, consider this: Working in smaller chunks compels you to think about the structure of your content more consciously. You are encouraged to divide your text into manageable parts that are easier for the chatbot not only to handle but also to streamline the review and editing process. The result is an efficient workflow that maintains high standards of output quality.
乍看之下，這似乎是一個額外的步驟。然而，請考慮這一點：以較小的部分進行工作迫使你更有意識地思考內容的結構。你被鼓勵將文本劃分為可管理的部分，這不僅使聊天機器人更容易處理，也使審查和編輯過程更為簡化。結果是一個高效的工作流程，保持高標準的輸出質量。

Often, we might feel the urge to provide as much information as possible to the AI, believing that more data will yield better results. However, the size limitation necessitates a different approach. It encourages us to be more selective, to remove redundant or peripheral information, and to focus on presenting the core content. This distillation of input can have a remarkable effect on the output, enhancing its relevance, precision, and overall quality.
我們常常可能會感到有必要向人工智慧提供盡可能多的信息，認為更多的數據會產生更好的結果。然而，大小限制需要我們採取不同的方法。這促使我們更加挑剔，去除冗餘或邊緣的信息，專注於呈現核心內容。這種對輸入的提煉可以對輸出產生顯著的影響，提高其相關性、精確性和整體質量。

In essence, these constraints act as a helpful filter. They push us to focus on delivering a concise and relevant input, which allows the chatbot to generate a more targeted and accurate output. This disciplined approach to interacting with AI can significantly improve the effectiveness of the AI models we work with, leading to better results in our tasks and projects.
本質上，這些限制充當了一個有用的過濾器。它們促使我們專注於提供簡潔且相關的輸入，這使得聊天機器人能夠生成更具針對性和準確性的輸出。這種與人工智慧互動的紀律性方法可以顯著提高我們所使用的人工智慧模型的有效性，從而在我們的任務和項目中獲得更好的結果。

Approach a longer text in stages. Start by writing down your ideas; there should be enough of them for a scholarly paper. Nothing can replace this initial
分階段處理較長的文本。首先寫下你的想法；應該有足夠的內容來撰寫學術論文。這一初步步驟是無法被替代的。

step, although, as I will show, AI can be used for brainstorming and preliminary research. Then ask the AI to turn your rough ideas into an outline, and complete with summaries for each proposed subheading. The AI’s unique capability of understanding the progression of a linear argument comes into play here, helping you to streamline and consolidate your thoughts. You will find that similar ideas that you might have considered as separate entities are gathered under one subheading, enhancing the clarity and coherence of your outline.
步驟，儘管如此，正如我將展示的，人工智慧可以用於頭腦風暴和初步研究。然後請人工智慧將您的粗略想法轉化為大綱，並為每個提議的副標題完成摘要。人工智慧獨特的理解線性論證進展的能力在這裡發揮作用，幫助您精簡和整合您的思路。您會發現，您可能認為是獨立實體的相似想法被歸納在一個副標題下，增強了您大綱的清晰度和一致性。

This strategic chunking ensures that your resulting content is well-paced. It prevents your narrative from feeling rushed or, conversely, from becoming too diluted. It ensures that your output maintains a balanced level of detail and flow, which is key to creating engaging, readable content.
這種策略性的分塊確保了您產出的內容節奏良好。它防止了您的敘述感覺匆忙，或者相反，變得過於稀釋。它確保您的輸出保持平衡的細節和流暢度，這是創造引人入勝、易於閱讀內容的關鍵。

This technique not only optimizes the AI’s text-processing capabilities but also supports you in constructing a well-structured, coherent narrative. In essence, this approach is a practical hack to navigate the size constraints of AI models and to leverage their capabilities for generating longer texts more efficiently.
這項技術不僅優化了人工智慧的文本處理能力，還幫助您構建一個結構良好、連貫的敘述。從本質上講，這種方法是一種實用的技巧，可以幫助您應對人工智慧模型的大小限制，並更有效地利用其生成更長文本的能力。

As previously discussed, there have been numerous attempts to expand both input and output sizes. Claude provides a direct method for inputting larger files, while ChatGPT can accommodate this through third-party plug-ins. However, overcoming the constraints on output size appears to be more challenging. Even if techniques are employed to coax the AI into producing a more expansive text, the quality seems to decline rapidly with the increase in size. It is likely that the necessity for dividing the content into manageable sections, or “chunking,” will remain a consideration for the foreseeable future.
如前所述，已經有許多嘗試擴大輸入和輸出大小。Claude 提供了一種直接的方法來輸入較大的文件，而 ChatGPT 則可以通過第三方插件來適應這一點。然而，克服輸出大小的限制似乎更具挑戰性。即使採用技術來引導 AI 產生更大範圍的文本，隨著大小的增加，質量似乎也會迅速下降。未來可預見的時間內，將內容劃分為可管理的部分或“分塊”的必要性可能仍然是一個考量。

1 Shakked Noy and Whitney Zhang, “Experimental Evidence on the Productivity Effects of Generative Artificial Intelligence,” available at SSRN 4375283 (2023), https://www. science.org/doi/10.1126/science.adh2586
1 Shakked Noy 和 Whitney Zhang，《生成性人工智慧的生產力影響的實驗證據》，可在 SSRN 4375283 獲得 (2023)，https://www.science.org/doi/10.1126/science.adh2586
2 Everett M. Rogers, Diffusion of Innovations (New York: The Free Press of Glencoe, 1962).
2 埃弗雷特·M·羅傑斯，《創新擴散》（紐約：格倫科自由出版社，1962 年）。

3 Alexander M. Sidorkin, “Tool Fetishism in Online Teaching,” 2020, https://sidorkin. blogspot.com/2020/08/tool-fetishism-in-online-teaching.html
3 亞歷山大·M·西多金，“線上教學中的工具迷戀，”2020 年，https://sidorkin。blogspot.com/2020/08/tool-fetishism-in-online-teaching.html
4 Kyle Wiggers, “Anthropic Releases Claude 2, Its Second-gen AI Chatbot,” Techcrunch, July 11, 2023, https://techcrunch.com/2023/07/11/anthropic-releases-claude-2-the-second-generation-of-its-ai-chatbot/
4 Kyle Wiggers，《Anthropic 發布 Claude 2，第二代 AI 聊天機器人》，Techcrunch，2023 年 7 月 11 日，https://techcrunch.com/2023/07/11/anthropic-releases-claude-2-the-second-generation-of-its-ai-chatbot/
5 OpenAI’s GPT-4. Chat, 7/22/2023.
5 OpenAI 的 GPT-4。聊天，2023 年 7 月 22 日。