AI是什么：理解人工智能的基本原理 | What Is AI: Understanding the Basics of Artificial Intelligence
What Is AI: Understanding the Basics of Artificial Intelligence
AI是什么：理解人工智能的基本原理
---
The Question That Started Everything: What Exactly Is AI?
一切的开端：AI究竟是什么？
When you ask your phone "What's the weather today?" and it responds with a forecast, or when Netflix suggests a show you actually want to watch, have you ever wondered what's happening behind the scenes? Is there a tiny person inside the machine, reading your request and thinking of an answer? The answer, of course, is no—but the reality is arguably even more fascinating.
当你对着手机问"今天天气怎么样"，它会给你一个天气预报；或者当你刷视频时，Netflix精准推荐了你正想看的内容——你有没有想过，这背后到底发生了什么？机器里是不是住着一个小人，正在读取你的问题、思考答案？当然不是。但真相往往比想象更加精彩。
Artificial Intelligence, commonly abbreviated as AI, refers to computer systems designed to perform tasks that typically require human intelligence. These tasks include understanding language, recognizing patterns, making decisions, and learning from experience. The key word here is "learning"—unlike traditional computer programs that follow rigid, predefined rules, AI systems can improve their performance over time by analyzing data and finding patterns on their own.
人工智能（Artificial Intelligence，简称AI）指的是那些被设计来执行通常需要人类智能才能完成的任务的计算机系统。这些任务包括理解语言、识别模式、做出决策，以及从经验中学习。这里的关键词是"学习"——与传统计算机程序遵循固定、预设的规则不同，AI系统能够通过分析数据、自主发现规律来提升自己的表现。
To truly understand AI, we need to trace its development history and explore how modern AI systems actually work. This chapter will take you on a journey from the origins of AI to the present day, from the theoretical foundations to hands-on experience. By the end, you'll be able to explain AI's basic working principles in your own words and identify at least three AI applications in your daily life.
要真正理解AI，我们需要追溯它的发展历程，探索现代AI系统究竟是如何运作的。本章将带你踏上一段从AI起源到当今时代的旅程，从理论基础到亲身实践。阅读完本章后，你将能够用自己的话解释AI的基本工作原理，并举出至少三个日常生活中的AI应用实例。
---
Section 1: The Evolution of AI—From the Turing Test to Large Language Models
第一节：AI的发展历程——从图灵测试到大语言模型
The Birth of an Idea: The Turing Test
概念的诞生：图灵测试
The concept of artificial intelligence has fascinated humans for centuries, but the formal birth of AI as a scientific discipline occurred in 1956, at a conference held at Dartmouth College in the United States. However, the foundations were laid years earlier by a British mathematician named Alan Turing.
人工智能的概念在几个世纪以来一直吸引着人类，但AI作为一门正式学科的诞生是在1956年，美国达特茅斯学院的一次会议上。不过，它的理论基础早在多年前就已奠定——贡献者是一位名叫艾伦·图灵的英国数学家。
In 1950, Alan Turing published a groundbreaking paper titled "Computing Machinery and Intelligence." In it, he posed a simple but profound question: "Can machines think?" To answer this question, he proposed what became known as the Turing Test—an experimental design where a human judge would converse with both a human and a machine through text-only communication, without knowing which was which. If the judge couldn't reliably tell the machine apart from the human, the machine was said to "pass" the test, implying it possessed human-level intelligence.
1950年，艾伦·图灵发表了一篇开创性的论文《计算机器与智能》。他在文中提出了一个简洁却深刻的问题："机器能思考吗？"为了回答这个问题，他设计了一个后来被称为图灵测试的实验：让一个人类评判者通过纯文本方式与一个人类和一个机器分别对话，且不知道哪个是哪个。如果评判者无法可靠地区分机器和人类，就说机器"通过"了测试——这意味着它具有人类水平的智能。
The Turing Test was revolutionary because it shifted the focus from how machines work internally to whether their behavior is indistinguishable from human behavior. This "behaviorist" approach became foundational to AI research for decades.
图灵测试的意义在于，它将关注点从机器内部如何运作转移到机器的行为是否与人类无法区分。这种"行为主义"方法在随后数十年里成为AI研究的基础。
Early Dreams and Disappointments
早期的梦想与挫折
The early years of AI research (1956-1974) were filled with optimism. Researchers believed that human-level AI was just around the corner. Government agencies, particularly the U.S. Department of Defense, poured millions of dollars into AI research. Simple programs were developed that could solve algebra problems, prove geometric theorems, and play simple games.
AI研究的早期岁月（1956-1974）充满了乐观情绪。研究人员相信，人类水平的AI近在咫尺。政府机构，尤其是美国国防部，向AI研究投入了数百万美元。人们开发出了能够解代数题、证明几何定理、下简单棋类的简单程序。
However, this enthusiasm soon met reality. The computing power of the time was extremely limited, and AI programs struggled with tasks that humans found easy—like understanding the real-world context of a simple sentence or recognizing objects in a cluttered image. By the mid-1970s, funding dried up, and the field entered its first "AI Winter."
然而，这种热情很快遭遇了现实的冷水。当时的计算能力极其有限，AI程序在处理人类觉得简单的任务时举步维艰——比如理解一个简单句子的真实世界语境，或在一张杂乱的图像中识别物体。到1970年代中期，资金枯竭，AI领域进入了第一个"AI寒冬"。
The lesson from this period is crucial: building intelligent machines is far more difficult than early researchers imagined. The human brain performs certain tasks so effortlessly that we take them for granted—until we try to program a computer to do them.
这个时期的教训至关重要：构建智能机器比早期研究者想象的要困难得多。 人脑执行某些任务是如此轻松，以至于我们习以为常——直到我们试图用计算机编程来实现它们时才发现其困难。
The Rise of Machine Learning
机器学习的崛起
After the AI Winter, the field gradually recovered with a shift in approach. Instead of trying to hard-code all the rules of intelligence, researchers began to develop systems that could learn from data. This approach, called Machine Learning (ML), became the dominant paradigm in AI research.
经历AI寒冬之后，这个领域以方法的转变逐渐复苏。研究人员不再试图硬编码所有智能规则，而是开始开发能够从数据中学习的系统。这种方法被称为机器学习（Machine Learning，简称ML），成为AI研究的主导范式。
Machine learning systems don't require explicit programming for every rule. Instead, they learn patterns from examples. For instance, to teach a system to recognize cats, you don't need to define "cat" with a list of rules (pointy ears, whiskers, etc.). Instead, you show the system thousands of pictures labeled "cat" and "not cat," and the system learns to distinguish between them.
机器学习系统不需要为每条规则编写显式程序。相反，它们从示例中学习。例如，要教会一个系统识别猫，你不需要用规则列表（尖耳朵、胡须等）来定义"猫"。相反，你向系统展示成千上万张标注为"猫"和"非猫"的图片，系统就会学会区分它们。
This data-driven approach proved remarkably successful. By the 2010s, machine learning had achieved superhuman performance in many narrow domains:
这种数据驱动的方法被证明非常成功。到2010年代，机器学习在许多狭义领域已经达到了超越人类的水平：
- **2011**: IBM's Watson defeated human champions on the quiz show *Jeopardy!*
- **2011**: IBM的Watson在益智问答节目《危险边缘》中击败了人类冠军
- **2016**: AlphaGo, developed by DeepMind, defeated the world champion in the ancient game of Go—a feat that experts had predicted would take decades to achieve
- **2016**: DeepMind开发的AlphaGo击败了围棋世界冠军——这是专家们曾预测需要几十年才能实现的壮举
- **2020s**: Large Language Models like GPT and Claude began demonstrating unprecedented abilities to understand and generate human-like text
- **2020年代**: 像GPT和Claude这样的大语言模型开始展现出前所未有的理解和生成类人文本的能力
The Era of Large Language Models
大语言模型时代
The most recent breakthrough in AI comes from Large Language Models (LLMs)—the technology behind conversational AI systems like the one you might be interacting with right now. LLMs represent a paradigm shift in how we build AI systems.
AI领域最新的突破来自大语言模型（Large Language Models，简称LLMs）——这就是你此刻可能正在与之交互的对话式AI背后的技术。LLMs代表了构建AI系统方式的范式转变。
Traditional AI systems were designed for specific tasks. A spam filter only detects spam; a chess program only plays chess. Large Language Models, by contrast, are trained on massive amounts of text data from the internet, books, and other sources. Through this training, they learn the patterns, structures, and relationships within human language—and they can apply this knowledge to an astonishing variety of tasks.
传统的AI系统是为特定任务设计的。垃圾邮件过滤器只检测垃圾邮件；象棋程序只下棋。相比之下，大语言模型是在来自互联网、书籍和其他来源的大量文本数据上进行训练的。通过这种训练，它们学习人类语言中的模式、结构和关系——并且可以将这些知识应用于各种各样的任务。
Modern LLMs can translate languages, write poetry, explain scientific concepts, write and debug computer code, and engage in detailed conversations on countless topics. This versatility is what makes them so powerful—and so important for everyone to understand, including young people who will interact with these systems throughout their lives.
现代LLM可以翻译语言、写诗、解释科学概念、编写和调试计算机代码，还能就无数话题进行详细对话。这种多功能性使它们变得如此强大——也让每个人都理解它们变得至关重要，包括将在整个生命中与这些系统互动的年轻人。
Understanding this evolution—from the Turing Test to today's Large Language Mode