什麼是AI Vibe編碼？

AI Vibe編碼是一種針對軟件開發的新方法，可以根據精心製作的自然語言用戶提示生成高級，可執行的代碼。

The term "vibe coding" was popularized by OpenAI co-founder Andrej Karpathy, who described it as a state of mind or attitude adopted by programmers when working with AI-generated code. It can be contrasted with the traditional approach to software development, which involves humans writing code from scratch in a structured and modular fashion, with each segment of code being carefully reviewed and signed off on by peers or superiors in a process known as "code inspection."

“ Vibe編碼”一詞由Openai聯合創始人Andrej Karpathy推廣，他將其描述為使用AI生成的代碼時採用程序員採用的心態或態度。它可以與傳統的軟件開發方法形成鮮明對比，該方法涉及人類以結構化和模塊化的方式從頭開始編寫代碼，每個代碼的每個部分都會在稱為“代碼檢查”的過程中仔細審查和簽署。

In essence, vibe coding is about letting the AI take the lead and generate the bulk of the code, while the human programmer acts more as a guide or supervisor, providing input and feedback along the way. This is not to say that no manual labor is involved; rather, it is minimal and focused on high-level tasks such as selecting the appropriate framework or setting the desired style for the output.

從本質上講，Vibe編碼是關於讓AI領先並生成大部分代碼，而人類程序員則更多地充當指導或主管，並在此過程中提供意見和反饋。這並不是說沒有涉及手動勞動。相反，它是最小的，並且專注於高級任務，例如選擇適當的框架或為輸出設置所需的樣式。

For instance, if a programmer wants to build a web application using React, they might input a command such as "generate a React component for displaying a list of items" and leave the rest to the AI. The AI would then generate the component code, which the programmer would glance at, perhaps making a small adjustment or two before moving on to the next task.

例如，如果程序員想使用React構建Web應用程序，則他們可能會輸入諸如“生成用於顯示項目列表的React組件”之類的命令，然後將其餘的留給AI。然後，AI將生成組件代碼，程序員會瞥了一眼，可能會在繼續執行下一個任務之前進行一兩個小調整。

This approach to software development is made possible by the recent advances in artificial intelligence, specifically in the area of natural language processing (NLP) and deep learning. Researchers have trained AI models to understand and generate multiple programming languages, enabling them to translate human instructions into executable code.

人工智能的最新進展，特別是自然語言處理（NLP）和深度學習方面的最新進展使這種軟件開發方法成為可能。研究人員已經培訓了AI模型，以理解和生成多種編程語言，從而使他們能夠將人類說明轉化為可執行的代碼。

One example of such a model is OpenAI's Codex, which is based on the same technology as the GPT-3 language model but trained on a massive dataset of code and text. Codex can perform various coding tasks, such as writing functions, translating code from one language to another, and even generating entire programs from scratch.

此類模型的一個示例是OpenAI的法典，該法典基於與GPT-3語言模型相同的技術，但在大量的代碼和文本數據集上進行了培訓。 Codex可以執行各種編碼任務，例如編寫功能，將代碼從一種語言轉換為另一種語言，甚至從頭開始生成整個程序。

Another AI model capable of vibe coding is DeepMind's AlphaCode, which was specifically designed for competitive programming, a domain that usually requires advanced algorithms and data structures. According to the researchers, AlphaCode's output is comparable to the performance of the median human competitor in a recent programming competition.

DeepMind的字母是專門為競爭性編程設計的另一個AI模型，該模型通常需要高級算法和數據結構。根據研究人員的說法，AlphaCode的輸出與最近的編程競賽中中位人類競爭對手的表現相當。

While AI vibe coding can lead to faster software development results, there is a risk that the output generated from a vibe coding approach may not be adequately reviewed and signed off on, which is where traditional coding practices continue to offer safeguards.

儘管AI Vibe編碼可以帶來更快的軟件開發結果，但可能不會對Vibe編碼方法產生的輸出產生的風險進行充分的審查和登錄，這是傳統的編碼實踐繼續提供保障措施的地方。

Moreover, if the programmer's input is not clear or consistent, the AI may generate code that deviates from the intended goal, leading to bugs or unexpected behavior. In this sense, AI is a tool that amplifies the programmer's intent and capabilities, but it does not eliminate the need for human supervision and feedback.

此外，如果程序員的輸入尚不清楚或不一致，則AI可能會生成偏離預期目標的代碼，從而導致錯誤或意外行為。從這個意義上講，AI是擴大程序員意圖和功能的工具，但並不能消除對人類監督和反饋的需求。

As AI technology continues to improve, we may see even more interesting and useful applications of vibe coding in the future, blurring the lines between human creativity and machine intelligence.

隨著AI技術的不斷改進，我們可能會在將來看到氛圍編碼的更有趣和有用的應用，從而模糊了人類創造力和機器智能之間的界限。