你好，AI朋友！——认识会思考的机器

第 1 章　你好，AI 朋友！——认识会思考的机器

Chapter 1: Hello, AI Friend! — Getting to Know Machines That Think

清晨六点半，闹钟还没响，小诺（Xiao Nuo）房间角落里的小音箱就亮起了一圈温柔的蓝光。它用清脆的声音说："早上好，小诺！今天星期三，杭州小雨转多云，气温 18 到 24 摄氏度，出门记得带一件薄外套哦。"小诺揉了揉眼睛，嘴里嘟囔了一句："谢谢！"然后他抬起手腕，智能手表轻轻震动了一下，屏幕上跳出一行字："该喝水啦！你昨天只喝了 4 杯，今天要补一杯哦！"

At 6:30 in the morning, before the alarm rang, the small speaker in Xiao Nuo's room lit up with a gentle blue glow. In a crisp voice it said, "Good morning, Xiao Nuo! Today is Wednesday, light rain turning cloudy in Hangzhou, 18 to 24 degrees Celsius. Don't forget to bring a light jacket when you go out."

小诺还没下床，就已经被三个"看不见的小帮手"照顾得妥妥帖帖。这个会说话、会关心人的小音箱，那个会数杯子的手表，还有放学后陪他猜谜语的聊天机器人——它们到底是什么？是人？是动物？还是某种魔法？这一章，我们就跟着小诺一起，去认识这位新朋友——人工智能（Artificial Intelligence，简称 AI）。

Before Xiao Nuo even got out of bed, three "invisible little helpers" had taken care of him. The speaking, caring speaker, the cup-counting watch, and the chatbot that plays riddles with him after school — what on earth are they? People? Animals? Some kind of magic? In this chapter, we'll follow Xiao Nuo to meet a new friend — Artificial Intelligence, or AI for short.

---

1.1　小诺的奇妙早晨：AI 就藏在你的生活里

1.1　Xiao Nuo's Magical Morning: AI Is Already Hiding in Your Life

让我们先做一件有趣的事情：把今天——从起床到睡觉——你遇到的所有"聪明机器"全都列出来。Vamos——让我们跟着小诺的一天来数一数：

Let's do something fun first: list every "smart machine" you meet today, from the moment you wake up until you fall asleep. Vamos — let's count along with Xiao Nuo's day:

• **早上起床**：智能音箱播报天气、智能手表提醒喝水、扫地机器人规划路线——这是 AI 在**听和说**。
• **上学路上**：爸爸用导航避开拥堵，路口的摄像头识别车牌——这是 AI 在**看和判断**。
• **课堂上**：平板上的学习 App 用语音帮你读英语单词，智能笔自动批改作业——这是 AI 在**教和评**。
• **回到家**：妈妈的手机自动把照片按"人物"和"风景"分类——这是 AI 在**识别和整理**。

• **Morning wake-up**: smart speaker reports weather, smartwatch reminds you to drink water, robot vacuum plans its route — that's AI **listening and speaking**.
• **On the way to school**: Dad uses navigation to avoid traffic, intersections have cameras that recognize license plates — that's AI **seeing and judging**.
• **In class**: the learning app on your tablet reads English words for you by voice, smart pens auto-grade homework — that's AI **teaching and evaluating**.
• **Back home**: Mom's phone automatically sorts photos by "people" and "scenery" — that's AI **recognizing and organizing**.

你发现了吗？AI 并不是住在某个遥远实验室里的神秘生物，它就在我们的手机、手表、音箱、汽车、玩具里，甚至就在你的身边。根据一份 2024 年的统计，全球已经有超过 30 亿人在每天使用 AI 工具——平均每两个人里，就有一个人在跟 AI 打交道。

Did you notice? AI isn't a mysterious creature living in some distant laboratory — it's in our phones, watches, speakers, cars, toys, even right beside you. According to a 2024 statistic, more than 3 billion people around the world use AI tools every day — on average, one out of every two people is interacting with AI.

那么问题来了：既然 AI 这么厉害，它到底是什么？它和爸爸妈妈电脑里那些不会说话的普通程序，有什么不一样？

So here's the question: if AI is so powerful, what exactly is it? How is it different from the silent, ordinary programs on your parents' computers?

---

1.2　像猜谜高手一样的电子小动物

1.2　Electronic Pets That Riddle Like Champions

如果让小诺用一句话告诉奶奶"AI 是什么"，他会怎么说呢？小诺眨巴着眼睛想了想，说："AI 就像一只特别特别聪明的电子小动物。它不会真的饿、不会真的困，但它能听懂我们说的话，还能回答我们的问题。"

If Xiao Nuo had to tell Grandma "what AI is" in one sentence, what would he say? Xiao Nuo blinked and thought for a moment: "AI is like a really, really smart electronic pet. It doesn't really get hungry or sleepy, but it can understand what we say and answer our questions."

这是一个很好的起点！我们顺着小诺的比喻，再往下走一步。在科学家眼里，今天最强大的一类 AI——也就是陪你聊天的机器人、写小作文的助手、翻译英语的工具——叫做大语言模型（Large Language Model，简称 LLM）。这个名字听起来吓人，但拆开来一点都不难：

This is a great starting point! Let's follow Xiao Nuo's metaphor one step further. In scientists' eyes, the most powerful type of AI today — the chatbots that chat with you, the assistants that write essays, the tools that translate English — is called a Large Language Model, or LLM for short. The name sounds scary, but broken down it's not difficult at all:

• **大（Large）**：训练这个 AI 用到的文字多到吓人。光是一种叫 GPT 的大语言模型，就"读"过几千亿句英文和中文句子。想象一下：你每天读 100 句话，要连续读 **3 万年** 才能读完那么多。
• **语言（Language）**：它学的不是魔法，而是**人类的语言**——中文、英文、日文、法文，甚至一点点古文和诗歌。
• **模型（Model）**：这个词来自数学家。模型就像一张"地图"，用很多很多**参数（Parameters）** 画出来。参数可以理解成 AI 脑子里的小旋钮——GPT 这样的模型里有 **几千亿个** 这样的小旋钮。

• **Large**: the amount of text used to train this AI is terrifyingly huge. One type of LLM called GPT has "read" hundreds of billions of English and Chinese sentences. Imagine: if you read 100 sentences every day, it would take you **30,000 years** of nonstop reading.
• **Language**: what it learns isn't magic — it's **human language** — Chinese, English, Japanese, French, even a little classical Chinese and poetry.
• **Model**: this word comes from mathematicians. A model is like a "map" drawn with countless **parameters**. You can think of parameters as little knobs in the AI's brain — a model like GPT has **hundreds of billions** of these little knobs.

> 💡 小词典 / Mini-Dictionary > > - 人工智能 / Artificial Intelligence (AI)：让机器做"原来只有人才能做"的事情，比如听懂话、认出猫、写故事。 > - 大语言模型 / Large Language Model (LLM)：用海量文字训练出来，专门擅长理解和生成语言的 AI。 > - 参数 / Parameters：AI 内部的"小旋钮"，通过调整它们，让 AI 的猜测变得越来越准。

> 💡 Mini-Dictionary > > - Artificial Intelligence (AI): getting machines to do "things only humans used to do," like understanding speech, recognizing cats, and writing stories. > - Large Language Model (LLM): an AI trained on huge amounts of text, especially good at understanding and generating language. > - Parameters: the "little knobs" inside an AI — by adjusting them, the AI's guesses become more and more accurate.

1.2.1　AI 真的在"思考"吗？

1.2.1　Is AI Really "Thinking"?

小诺又问了一个非常聪明的问题："那 AI 真的像我们一样，会思考吗？"科学家会老老实实地告诉你：不完全是。 AI 没有真正的意识，它不会饿、不会累、不会想念奶奶。但它能做出一些看起来很像"思考"的事情。这是为什么呢？

Xiao Nuo asked another very clever question: "Does AI really think, the way we do?" Scientists will honestly tell you: not entirely. AI has no real consciousness — it doesn't get hungry, tired, or miss Grandma. But it can do things that look a lot like "thinking." Why is that?

我们来看 AI 做的一个最基础的动作——猜下一个字。

Let's look at one of the most basic things AI does — guessing the next word.

请你试试看：如果你看到一句话的开头是"今天天气真___"，你的脑子里最先冒出来的是什么字？很多人会想到"好"，也有人会想到"冷"、"热"、"不错"。这种根据上下文来"猜下一个字"的动作，其实就是大语言模型最核心的工作方式。

Try this: if you see a sentence that starts with "Today the weather is really ___", what's the first word that pops into your head? Many people think "nice," others think "cold," "hot," "great." This action of guessing the next word based on context is actually the core way a large language model works.

> 🎲 小实验 / Mini-Experiment > > 请你拿起笔，写下三句话，每句留一个空，让爸爸妈妈、爷爷奶奶、好朋友来填： > > Pick up your pen and write three sentences, each with one blank, and let your parents, grandparents, and best friend fill them in: > > 1. 我最喜欢吃的水果是 ___。 > 2. 小狗最害怕的事情是 ___。 > 3. 长大以后，我想当 ___。 > > 注意：每个人填的字可能都不一样！这就是语言的有趣之处——同一个开头，可以接不同的字。AI 也很擅长做这件事，只是它"读过的句子"比我们多得多。

> 🎲 Mini-Experiment > > Pick up your pen and write three sentences, each with a blank, and ask your family and friends to fill them in. Notice: everyone might fill in different words! That's the fun of language — the same opening can lead to different words. AI is also great at this, except it has "read" far, far more sentences than we have.

1.2.2　上下文窗口：AI 的"短期记忆"

1.2.2　The Context Window: AI's "Short-Term Memory"

AI 虽然能读很多书，但它有一个有趣的限制——它一次能"看到"的文字是有限的。科学家把 AI 一次能处理的文字长度叫做"上下文窗口（Context Window）"。你可以把它想象成 AI 的"短期记忆"或者一张"草稿纸"。

Although AI can read many books, it has an interesting limitation — the amount of text it can "see" at one time is limited. Scientists call the length of text AI can process at one time the "context window." You can imagine it as AI's "short-term memory" or a piece of "scratch paper."

打个比方：你和同学讨论一部很长的电影。刚开始讨论的部分你记得清清楚楚，但如果讨论得太久，你可能会忘记 20 分钟前说过的话。AI 也是这样——如果对话太长，它会"忘记"最早的内容。所以当你和 AI 聊天时，把最重要的事情放在对话的开头或者最近的位置，往往会得到更好的答案。

Let's make a comparison: you and your classmates are discussing a very long movie. You remember the parts you discussed at