📚 IB & Edexcel Computer Science: AI Core Revision Notes | IB与爱德思计算机：人工智能考点精讲
Artificial Intelligence (AI) sits at the heart of modern computer science and is a key topic examined in both IB Diploma and Edexcel A‑Level specifications. This revision guide distils the essential concepts – from definitions and search algorithms to machine learning paradigms and ethical considerations – aligning with the assessment objectives of both curricula.
人工智能是当代计算机科学的核心，也是 IB 文凭和爱德思 A‑Level 两大课程体系的必考主题。本文精讲关键考点，涵盖定义、搜索算法、机器学习范式以及伦理影响，助你高效备考，紧扣两大考纲的评估目标。
1. What is Artificial Intelligence? | 什么是人工智能？
AI refers to the development of computer systems that can perform tasks normally requiring human intelligence, such as reasoning, learning, perception and natural language understanding. It is broadly divided into narrow AI, which excels at specific tasks, and general AI, a theoretical system with human‑like cognitive abilities across domains.
人工智能指开发能执行通常需要人类智能才能完成的任务的计算机系统，这些任务包括推理、学习、感知和自然语言理解。AI 大致分为擅长特定任务的弱人工智能，以及理论上具备跨领域人类认知能力的通用人工智能。
In both IB and Edexcel syllabi, students must distinguish between strong AI (with consciousness) and weak AI (simulated intelligence). Many real‑world applications, such as game playing, voice assistants and recommendation engines, are examples of weak AI programming.
IB 与爱德思考纲均要求考生能区分强人工智能（具备意识）和弱人工智能（模拟智能）。棋类游戏、语音助手和推荐引擎等大量现实应用都属于弱人工智能程序。
The concept of rationality is also central: an AI agent acts rationally if it selects actions that maximise its performance measure given its knowledge and percepts.
理性概念同样关键：若一个智能体根据其知识和感知选择行动以最大化性能指标，则其行为是理性的。
2. The Turing Test and Intelligent Behaviour | 图灵测试与智能行为
Alan Turing proposed a practical test for machine intelligence: a human interrogator engages in natural language conversations with a machine and another human via text. If the interrogator cannot reliably distinguish the machine from the human, the machine is said to have passed the test.
艾伦·图灵提出了检验机器智能的实用方法：人类询问者通过文本与一台机器和另一名人类进行自然语言对话。如果询问者无法可靠地区分出机器，则该机器被认为通过了测试。
The Turing Test evaluates behaviour, not internal mechanisms. Critics argue that passing the test does not prove true understanding – the Chinese Room argument demonstrates that symbol manipulation alone can create the illusion of intelligence without comprehension.
图灵测试评估的是行为而非内在机制。批评者指出，通过测试并不证明真正理解——“中文屋”思想实验表明，纯符号操作可以制造智能假象，却无真正理解。
Both IB and Edexcel examinations may ask candidates to discuss the strengths and limitations of the Turing Test as a measure of AI, connecting it to topics like natural language processing and philosophical questions of mind.
IB 和爱德思考试均可能要求考生讨论图灵测试作为人工智能衡量标准的优势与局限，并将其联系到自然语言处理及心灵哲学问题。
3. Problem Solving and Search Algorithms | 问题求解与搜索算法
AI problem solving often involves exploring a state space to find a sequence of actions leading from an initial state to a goal state. Search algorithms are fundamental in games, route planning and puzzle solving.
人工智能问题求解通常需要探索状态空间，以寻找从初始状态到达目标状态的动作序列。搜索算法在游戏、路径规划和谜题求解中至关重要。
Uninformed search algorithms, such as breadth‑first search (BFS) and depth‑first search (DFS), be explored systematically. BFS guarantees the shortest path in an unweighted graph, while DFS is memory‑efficient. Performance is evaluated by completeness, optimality, time and space complexity.
无信息搜索算法如广度优先和深度优先搜索，系统探索状态空间。BFS 保证在无权图中找到最短路径，DFS 则内存效率高。算法性能从完备性、最优性、时间复杂度和空间复杂度四个维度评估。
Informed (heuristic) search uses domain knowledge to guide exploration. A* search combines the exact cost g(n) with a heuristic estimate h(n): f(n) = g(n) + h(n). A* is optimal when the heuristic is admissible – it never overestimates the true cost.
有信息（启发式）搜索利用领域知识引导探索。A* 算法结合实际代价 g(n) 与启发式估计 h(n)：f(n) = g(n) + h(n)。当启发式函数是可采纳的（从不高于真实代价），A* 算法能保证最优性。
A* evaluation function: f(n) = g(n) + h(n)
Edexcel especially expects candidates to trace search algorithms on simple graphs and compare their efficiency. IB Higher Level may require understanding of minimax and alpha‑beta pruning for adversarial search.
爱德思考纲尤其要求考生能够在简单图上模拟搜索算法并比较效率。IB 高水平可能要求理解对抗搜索中的极小极大算法和 alpha‑beta 剪枝。
4. Knowledge Representation and Reasoning | 知识表示与推理
For an AI to reason about the world, knowledge must be represented in a formal structure. Common approaches include semantic networks, frames, rules and logic.
为使人工智能能对世界进行推理，知识必须以形式化结构表示。常用方法包括语义网络、框架、规则和逻辑。
Rule‑based systems express knowledge as IF‑THEN rules. A forward chaining inference engine starts from known facts and applies rules to reach a conclusion; backward chaining starts from a goal and works backwards. These are examinable in both syllabi, often within expert systems.
基于规则的系统用 IF‑THEN 规则表达知识。正向链推理引擎从已知事实出发应用规则得出结论；反向链则从目标出发逆向推理。两种链式推理在两套课程中均是考点，常出现在专家系统语境中。
Propositional and predicate logic enable rigorous reasoning. For example, first‑order logic can represent statements like ∀x Cat(x) → Mammal(x). Students should understand how inference engines use resolution to derive new facts.
命题逻辑和谓词逻辑支持严格推理。例如，一阶谓词逻辑可表达“所有猫都是哺乳动物”：∀x Cat(x) → Mammal(x)。考生需理解推理引擎如何利用归结原理推导新事实。
| Forward Chaining | Backward Chaining |
|---|---|
| Data‑driven, works from initial facts | Goal‑driven, works from hypothesis |
| Good for monitoring and diagnosis | Effective when goal is known |
5. Machine Learning: Supervised vs. Unsupervised | 机器学习：有监督与无监督
Machine learning is a core AI technique where systems improve performance through experience. The two primary paradigms tested are supervised and unsupervised learning.
机器学习是人工智能的核心技术，系统可通过经验提升性能。考试中涉及的两大基本范式是有监督学习与无监督学习。
Supervised learning uses labelled datasets where each input comes with a correct output. The algorithm learns a mapping function to predict outputs for new data. Classification (discrete labels) and regression (continuous values) are typical tasks. Linear regression, decision trees and support vector machines are common models.
有监督学习使用带标签的数据集，每个输入都配有正确的输出。算法学习一个映射函数以对新数据输出预测。分类（离散标签）和回归（连续值）是典型任务；常用模型有线性回归、决策树和支持向量机。
Unsupervised learning works with unlabelled data to uncover hidden patterns. Clustering (e.g., k‑means) groups similar data points; association rule mining discovers relationships between items. Dimensionality reduction techniques like PCA are also examined.
无监督学习在无标签数据上寻找隐藏模式。聚类（如 k‑均值）将相似数据点归组；关联规则挖掘发现项目间的关系。主成分分析等降维技术也是考查内容。
The IB syllabus may also include semi‑supervised and reinforcement learning, while Edexcel emphasises the importance of training/test splits and overfitting prevention.
IB 大纲可能还涉及半监督和强化学习，爱德思则强调训练集/测试集划分及防止过拟合的重要性。
6. Reinforcement Learning and Deep Learning | 强化学习与深度学习
Reinforcement learning (RL) is a paradigm where an agent learns to make decisions by interacting with an environment. The agent receives rewards or penalties based on its actions and aims to maximise cumulative reward. Key components include policy, reward signal, value function and often a model of the environment.
强化学习是一种让智能体通过与环境的交互学会决策的范式。智能体依据行动获得奖励或惩罚，目标是最大化累积奖励。关键要素包括策略、奖励信号、价值函数，有时还包括环境模型。
Markov Decision Processes (MDPs) provide a mathematical framework for RL problems. Q‑learning is a model‑free technique that learns action‑value functions without needing a transition model. It is referenced in IB HL AI options.
马尔可夫决策过程（MDP）为强化学习提供数学框架。Q 学习是一种无模型技术，无需转移模型即可学习动作价值函数，属于 IB 高水平 AI 选修内容。
Deep learning uses multi‑layered artificial neural networks to model complex patterns. Deep neural networks automatically learn hierarchical feature representations from raw data, revolutionising fields like image and speech recognition. Both IB and Edexcel require awareness of its transformative impact, though detailed implementation is not required.
深度学习利用多层人工神经网络对复杂模式建模。深度网络能自动从原始数据中学习分层特征表示，彻底变革了图像和语音识别等领域。IB 与爱德思均要求了解其变革性影响，但不涉及具体实现细节。
7. Neural Networks and Perceptrons | 神经网络与感知器
The perceptron is the simplest artificial neural unit. It computes a weighted sum of inputs plus a bias and applies an activation function. If the sum exceeds a threshold, the perceptron fires.
感知器是最简单的人工神经单元。它计算输入加权和加上偏置，再经激活函数处理。若加权和超过阈值，感知器则激活。
y = step(∑ wᵢxᵢ + b), where step(z) = 1 if z ≥ 0 else 0
A single perceptron can only solve linearly separable problems; the XOR problem is a classic limitation. Multilayer perceptrons with non‑linear activation functions (e.g., sigmoid, ReLU) overcome this by forming decision boundaries that are non‑linear.
单感知器只能解决线性可分问题，XOR 问题就是其经典局限。采用非线性激活函数（如 Sigmoid、ReLU）的多层感知器可通过生成非线性决策边界克服这一限制。
Backpropagation is the essential algorithm for training neural networks. It calculates the gradient of the loss function with respect to each weight by applying the chain rule, allowing gradient descent to update weights iteratively.
反向传播是训练神经网络的核心算法。它通过链式法则计算损失函数对各权重的梯度，使梯度下降法能够迭代更新权重。
IB Higher Level and Edexcel frequently examine the structure of a basic neural network, the role of weights and biases, and the principle of gradient descent.
IB 高水平与爱德思考纲频繁考查基础神经网络的结构、权重与偏置的作用，以及梯度下降的原理。
8. Natural Language Processing and Computer Vision | 自然语言处理与计算机视觉
Natural Language Processing (NLP) enables machines to understand, interpret and generate human language. Key stages include tokenisation, stemming/lemmatisation, part‑of‑speech tagging and semantic analysis. Syntax trees and context‑free grammars are often examined for parsing.
自然语言处理（NLP）使机器能够理解、解释和生成人类语言。关键步骤包括分词、词干提取/词形还原、词性标注和语义分析。句法树和上下文无关文法常出现在语法分析考题中。
Computer Vision deals with enabling computers to gain high‑level understanding from digital images or videos. Image segmentation, feature extraction and convolutional neural networks (CNNs) are important concepts. Edge detection using filters like Sobel is a classic example.
计算机视觉致力于使计算机从数字图像或视频中获取高层理解。图像分割、特征提取和卷积神经网络（CNN）是重要概念。使用 Sobel 等滤波器进行边缘检测是经典实例。
Both fields illustrate modern AI applications heavily reliant on machine learning. Edexcel may include case studies on chatbots and facial recognition, while IB may explore the Turing Test’s relevance to NLP.
这两个领域展示了现代人工智能对机器学习的严重依赖。爱德思可能涉及聊天机器人和人脸识别的案例研究；IB 则可能探讨图灵测试与 NLP 的相关性。
9. Ethical and Social Implications of AI | 人工智能的伦理与社会影响
AI raises profound ethical questions concerning bias, privacy, accountability and employment. Algorithmic bias can arise from skewed training data, leading to unfair outcomes in hiring, credit scoring or criminal justice.
人工智能引发了关于偏见、隐私、问责和就业的深层伦理问题。算法偏见可能源自不均衡的训练数据，导致招聘、信用评分或刑事司法中的不公结果。
Autonomous decision‑making, as in self‑driving cars, forces us to consider the ‘trolley problem’ and legal responsibility. Transparency and explainability of AI models – particularly black‑box deep learning – are crucial concerns for both IB TOK links and Edexcel long‑answer questions.
自动驾驶汽车中的自主决策迫使我们思考“电车难题”及法律责任。AI 模型的透明度与可解释性（尤其是黑箱深度学习）是 IB 知识论联系以及爱德思长篇问答中的关键关注点。
Privacy issues emerge from mass data collection for training AI; surveillance capitalism and data ownership are hotly debated. Students should discuss the importance of data protection regulations like GDPR.
大规模数据收集用于训练 AI 导致隐私问题；监控资本主义和数据所有权备受争议。学生应讨论 GDPR 等数据保护法规的重要性。
Additionally, the impact of automation on employment and the digital divide must be weighed. AI benefits should be distributed equitably to avoid exacerbating societal inequalities.
此外，自动化对就业的影响以及数字鸿沟也需要权衡。人工智能带来的利益应公平分配，以免加剧社会不平等。
10. AI Applications and Future Trends | 人工智能应用与未来趋势
AI is embedded in everyday technologies: recommendation systems on streaming platforms, smart assistants, fraud detection in banking, and predictive text. Understanding these applications helps contextualise theoretical concepts.
人工智能已嵌入日常技术：流媒体平台的推荐系统、智能助手、银行欺诈检测以及预测文本。理解这些应用有助于将理论概念置于实际背景中。
In healthcare, AI aids diagnostic imaging and drug discovery; in transportation, it powers autonomous navigation. Both IB and Edexcel value the ability to discuss real‑world case studies and evaluate their effectiveness.
在医疗领域，AI 辅助诊断成像和药物发现；在交通领域，它驱动自主导航。IB 与爱德思均重视讨论真实案例并评估其有效性的能力。
Future trends include explainable AI (XAI), edge AI running on local devices, and the convergence of AI with the Internet of Things (AIoT). The development of artificial general intelligence remains a long‑term aspiration but raises significant scientific and ethical challenges.
未来趋势包括可解释人工智能（XAI）、在本地设备上运行的边缘 AI 以及与物联网融合的 AIoT。通用人工智能的研发仍是长期目标，但带来重大科学和伦理挑战。
Exam questions may ask for evaluations of current AI capabilities versus human intelligence, or for predictions about AI’s societal impact, drawing on all the topics covered in this guide.
考试题目可能要求评估当前 AI 能力与人类智能的对比，或预测 AI 的社会影响，需要综合运用本指南涵盖的所有主题。
Published by TutorHao | Computer Science Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
屏轩国际教育cambridge primary/secondary checkpoint, cat4, ukiset,ukcat,igcse,alevel,PAT,STEP,MAT, ibdp,ap,ssat,sat,sat2课程辅导，国外大学本科硕士研究生博士课程论文辅导