📚 AI Revision for IB & AQA Computer Science | 人工智能考点精讲
Artificial Intelligence (AI) is a rapidly evolving field that forms a key part of both the IB and AQA Computer Science syllabi. This article unpicks the essential concepts, from foundational definitions to ethical dilemmas, helping you master the topic for exams. Whether you are revising the Turing Test, neural networks, or the social impact of automation, this guide provides clear explanations paired with bilingual commentary to reinforce your understanding.
人工智能是 IB 与 AQA 计算机科学考纲中快速发展的核心领域。本文拆解从基础定义到伦理困境的关键概念，帮助你攻克考试。无论你正在复习图灵测试、神经网络还是自动化的社会影响，本指南都提供清晰的双语对照解析，强化理解。
1. What is Artificial Intelligence? | 什么是人工智能？
AI refers to the simulation of human intelligence in machines that are programmed to think and learn. In exam terms, it is often broken down into the ability to reason, discover meaning, generalise, and learn from experience. Both IB and AQA emphasise the distinction between simply executing a program and truly 'intelligent' behaviour.
人工智能指在机器上模拟人类智能，使其具备思考与学习能力。在考试中，它常被拆分为推理、发现意义、泛化以及从经验中学习的能力。IB 和 AQA 均强调单纯执行程序与真正“智能”行为之间的区别。
- AI systems can perceive their environment and take actions to maximise the chance of success.
- 人工智能系统能够感知环境并采取行动以最大化成功机会。
- The Turing Test, proposed by Alan Turing, evaluates a machine's ability to exhibit intelligent behaviour indistinguishable from a human.
- 图灵测试由艾伦·图灵提出，评估机器是否表现出与人类无异的智能行为。
- Strong AI vs Weak AI: Weak AI is designed for a specific task (e.g., chess), while Strong AI would have generalised human cognitive abilities. (See Topic 2).
- 强人工智能与弱人工智能：弱人工智能专为特定任务设计（如下棋），强人工智能则具备通用的人类认知能力（见第2节）。
2. Types of AI: Narrow vs General | 人工智能的类型：弱人工智能与强人工智能
Narrow AI (or weak AI) operates within a limited pre-defined range. All current AI systems, from voice assistants to recommendation engines, fall into this category. General AI, which remains theoretical, would possess the ability to understand, learn, and apply knowledge across different domains, much like a human. The IB syllabus expects you to differentiate these with clear examples.
弱人工智能在预设的有限范围内运作。从语音助手到推荐引擎，当前所有 AI 系统均属此列。强人工智能目前仍是理论概念，它能跨领域理解、学习和应用知识，类似于人类。IB 考纲要求你用清晰的例子区分两者。
For AQA, candidates must also understand that narrow AI is often rule-based or data-driven and lacks consciousness or self-awareness. This distinction is crucial when discussing AI limitations in ethical scenarios.
对 AQA 考生而言，还必须理解弱人工智能通常基于规则或数据驱动，缺乏意识或自我认知。在讨论伦理情境中 AI 的局限性时，这一区别至关重要。
| Narrow AI | General AI |
| Performs specific tasks (e.g., facial recognition) | Handles any intellectual task a human can |
| No true understanding; relies on patterns | Would possess consciousness (hypothetical) |
| Examples: Siri, AlphaGo, spam filters | No real-world examples exist yet |
3. Machine Learning Basics | 机器学习基础
Machine Learning (ML) is a subset of AI where systems improve their performance on a task through experience (data) without being explicitly programmed. The algorithm builds a mathematical model based on sample data, known as training data, to make predictions or decisions. Key concepts include training sets, test sets, features, and labels.
机器学习是 AI 的子集，系统通过经验（数据）而非显式编程来提升任务表现。算法基于样本数据（即训练数据）构建数学模型，用于预测或决策。关键概念包括训练集、测试集、特征和标签。
A common example is a spam filter: the ML model learns from thousands of emails labelled 'spam' or 'not spam' and then applies that learning to new emails. In the IB syllabus, you may be asked to describe how a model is trained and tested, including the role of the test set in avoiding overfitting.
常见示例是垃圾邮件过滤器：ML 模型从成千上万封标记为“垃圾邮件”或“非垃圾邮件”的邮件中学习，然后将学习成果应用到新邮件上。IB 考纲可能要求你描述模型的训练和测试过程，包括测试集在避免过拟合中的作用。
4. Supervised vs Unsupervised Learning | 监督学习与无监督学习
Supervised learning uses labelled data to train models. The algorithm learns a mapping from inputs to outputs. Common algorithms include linear regression (for continuous values) and decision trees (for classification). For example, predicting house prices based on features like size and location is supervised learning.
监督学习使用带标签的数据训练模型。算法学习输入到输出的映射。常见算法包括线性回归（用于连续值）和决策树（用于分类）。例如，根据面积和位置等特征预测房价就是监督学习。
Unsupervised learning works with unlabelled data, trying to find hidden patterns or structures. Clustering (e.g., k-means) groups similar data points without prior knowledge of categories. An example is customer segmentation in marketing. The AQA specification expects recognition of these learning types and real-world applications.
无监督学习处理无标签数据，试图发现隐藏的模式或结构。聚类（如 k-均值）在无先验类别知识的情况下将相似数据点分组。市场营销中的客户细分就是其一。AQA 考纲要求识别这些学习类型及其实际应用。
Supervised → Labelled data + desired output | 监督学习 → 有标签数据 + 期望输出
Unsupervised → No labels; discovers patterns | 无监督学习 → 无标签；发现模式
5. Neural Networks and Deep Learning | 神经网络与深度学习
Neural networks are inspired by the human brain and consist of layers of interconnected nodes (neurons). Each connection has a weight that adjusts during training. The input layer receives data, hidden layers process it, and the output layer produces the result. Activation functions (like ReLU or sigmoid) introduce non-linearity, allowing the network to learn complex patterns.
神经网络受人类大脑启发，由相互连接的节点（神经元）层组成。每个连接有权重，在训练中调整。输入层接收数据，隐藏层处理数据，输出层产生结果。激活函数（如 ReLU 或 sigmoid）引入非线性，使网络能学习复杂模式。
Deep learning refers to neural networks with many hidden layers. This depth enables handling of unstructured data such as images, audio, and natural language. Exam questions may ask you to explain the role of backpropagation — the algorithm that computes gradients to update weights and minimise loss. Both IB and AQA highlight the importance of training data quantity and quality for deep learning success.
深度学习指具有许多隐藏层的神经网络。这种深度能够处理图像、音频和自然语言等非结构化数据。考题可能要求你解释反向传播的作用——该算法计算梯度以更新权重并最小化损失。IB 和 AQA 都强调训练数据的数量与质量对深度学习成功的重要性。
6. Natural Language Processing (NLP) | 自然语言处理
NLP involves the interaction between computers and human language. It enables machines to understand, interpret, and generate text. Core tasks include sentiment analysis (determining emotion), named entity recognition (extracting names/places), and machine translation. Techniques like tokenisation, stemming, and part-of-speech tagging preprocess text for analysis.
自然语言处理涉及计算机与人类语言的交互。它使机器能理解、解释和生成文本。核心任务包括情感分析（判断情绪）、命名实体识别（提取名称/地点）和机器翻译。分词、词干提取和词性标注等技术预先处理文本以供分析。
Modern NLP often uses transformer architectures (e.g., BERT, GPT) that capture context better than previous models. The AQA syllabus touches on how chatbots and voice assistants utilise NLP to interpret user intent, while IB may ask about the ethical implications of language models generating misleading information.
现代 NLP 常使用 Transformer 架构（如 BERT、GPT），比以往模型更好地捕捉上下文。AQA 考纲涉及聊天机器人和语音助手如何利用 NLP 解读用户意图，而 IB 可能会问语言模型生成误导信息的伦理影响。
7. Computer Vision | 计算机视觉
Computer vision enables machines to interpret and make decisions based on visual data. Convolutional neural networks (CNNs) are the backbone of many vision systems, applying filters to detect edges, textures, and objects. Successive layers extract increasingly abstract features. Applications include facial recognition, autonomous vehicles, and medical imaging diagnosis.
计算机视觉使机器能够基于视觉数据进行解读与决策。卷积神经网络是许多视觉系统的支柱，通过滤波器检测边缘、纹理和物体。后续层提取越来越抽象的特征。应用包括人脸识别、自动驾驶车辆和医学影像诊断。
Exam tips: be prepared to describe how a CNN convolution layer works — an image patch multiplies element-wise with a kernel and sums to produce a feature map. Also, discuss limitations such as adversarial attacks where slight pixel changes fool the model. AQA frequently includes such cybersecurity-related AI topics.
考试技巧：准备好描述 CNN 卷积层如何工作——图像块与卷积核逐元素相乘并求和，生成特征图。同时讨论局限性，如对抗性攻击，轻微像素变化即可欺骗模型。AQA 经常包含此类网络安全相关的 AI 议题。
8. Expert Systems | 专家系统
An expert system is an AI program that emulates the decision-making ability of a human expert. It uses a knowledge base (facts and rules) and an inference engine to derive conclusions. They were popular in early AI and remain relevant for diagnostic tasks (e.g., medical diagnosis, fault detection).
专家系统是模拟人类专家决策能力的 AI 程序。它利用知识库（事实与规则）和推理引擎得出结论。这类系统在早期 AI 中很流行，如今在诊断任务（如医疗诊断、故障检测）中仍具相关性。
Key components: the knowledge base stores domain-specific information, often as IF-THEN rules; the inference engine applies logical reasoning (forward or backward chaining) to solve problems. The IB syllabus expects you to visualise a simple rule-based system and understand its limitations — brittle when facing scenarios outside predefined rules.
关键组件：知识库存储领域信息，通常采用 IF-THEN 规则；推理引擎应用逻辑推理（正向或反向链）来解决问题。IB 考纲要求你能想象简单的基于规则的系统，并理解其局限——面对超出预设规则的情形时脆弱易损。
9. Ethics and Bias in AI | 人工智能中的伦理与偏见
AI systems can inherit and amplify biases present in training data, leading to unfair outcomes in hiring, lending, and criminal justice. Both boards stress the ethical considerations: transparency, accountability, privacy, and the impact on employment. The European Union's AI Act or similar regulations often appear in exam discussions.
AI 系统会继承并放大训练数据中的偏见，导致招聘、贷款和刑事司法等领域的不公平结果。两大考局均强调伦理考量：透明度、问责制、隐私以及对就业的影响。欧盟《人工智能法案》等法规常出现在讨论中。
Students must evaluate the balance between innovation and regulation. Concepts like algorithmic fairness, explainable AI (XAI), and data protection (GDPR) are essential. When answering exam questions, use structured arguments: identify the ethical dilemma, discuss stakeholders, and suggest mitigations like bias audits or diverse datasets.
学生必须评估创新与监管之间的平衡。算法公平性、可解释 AI 和数据保护等概念必不可少。作答时，使用结构化论证：指明伦理困境、讨论利益相关方，并提出缓解措施，如偏见审计或多样化数据集。
10. AI Applications and Future | 人工智能的应用与未来
Beyond the classroom, AI powers healthcare (drug discovery, personalised medicine), finance (fraud detection, algorithmic trading), and transportation (autonomous drones, traffic optimisation). The exam may ask you to link an application to the underlying AI technique — e.g., reinforcement learning for game playing or generative adversarial networks (GANs) for creating art.
课堂之外，AI 在医疗（药物发现、个性化医疗）、金融（欺诈检测、算法交易）和交通（自动驾驶无人机、交通优化）等领域发挥重要作用。考试可能要求你将应用与底层 AI 技术联系起来——例如，用强化学习玩游戏或用生成对抗网络创作艺术。
The future of AI involves challenges like general intelligence, energy-efficient computing, and human-AI collaboration. IB questions may invite you to discuss the societal readiness for autonomous systems and the need for lifelong learning. AQA may include case studies on the digital divide and how AI can widen or narrow it.
AI 的未来包含通用智能、节能计算和人机协作等挑战。IB 题目可能邀请你讨论社会对自主系统的接受度以及终身学习的必要性。AQA 可能包含有关数字鸿沟的案例研究，探讨 AI 如何扩大或缩小鸿沟。
11. Exam Tips for IB and AQA | IB与AQA考试技巧
For IB Paper 1 (Case Study) and Paper 2 (Core Concepts), AI topics appear in both short-answer and essay questions. Define terms precisely (e.g., 'machine learning is a subset of AI that…'), give concrete examples, and always evaluate — 'while AI improves efficiency, it may also introduce bias.' AQA's Paper 2 focuses on ethical, legal, and environmental impacts, so revise those.
IB 试卷一（案例研究）和试卷二（核心概念）中，AI 题目出现在简答与论述中。精准定义术语（如“机器学习是 AI 的子集，它……”），给出具体例子，并始终进行评估——“虽然 AI 提高了效率，但也可能引入偏见。” AQA 试卷二侧重伦理、法律和环境影响，需重点复习。
- Use diagrams: for neural networks, draw input-hidden-output layers with labels.
- 使用图表：画神经网络时，标注输入-隐藏-输出层。
- Link topics: how IoT data feeds into ML models to create smart cities.
- 联系话题：物联网数据如何输入 ML 模型以创建智慧城市。
- Stay updated: exam boards love recent examples like AI in vaccine development.
- 紧跟时事：考局偏爱疫苗研发中的 AI 等最新例子。
- Manage time: bullet points in longer answers can save time; but write in full sentences where required.
- 时间管理：长答案中使用项目符号可节省时间；按要求使用完整句子。
Published by TutorHao | AI Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图