📚 A-Level AQA Computer Science: Artificial Intelligence Revision Notes | A-Level AQA 计算机：人工智能考点精讲
Artificial Intelligence (AI) is one of the most exciting and rapidly evolving fields in computer science. For AQA A-Level Computer Science, the AI option examines how machines can be designed to exhibit intelligent behaviour, covering core concepts such as intelligent agents, search algorithms, knowledge representation, machine learning, and the wider ethical implications. This article distils exactly what you need to know for the exam, with clear English–Chinese paired explanations and structured revision notes.
人工智能是计算机科学中最激动人心、发展最快的领域之一。在 AQA A-Level 计算机科学中，人工智能选修模块探讨如何设计表现出智能行为的机器，涵盖智能体、搜索算法、知识表示、机器学习等核心概念，以及更广泛的伦理影响。本文提炼了考试所需的所有要点，采用英中对照的清晰解释和结构化复习笔记。
1. What is Artificial Intelligence? | 什么是人工智能？
Artificial Intelligence is the science and engineering of making machines, especially computer programs, that can perform tasks requiring human-like intelligence. These tasks include learning from experience, reasoning, problem-solving, perception, and language understanding. A common distinction is made between weak AI (narrow AI), which is designed to perform a specific task, and strong AI (general AI), which would possess the full range of human cognitive abilities.
人工智能是一门科学与工程，旨在制造能够执行需要类似人类智能的任务的机器，尤其是计算机程序。这些任务包括从经验中学习、推理、解决问题、感知和语言理解。通常区分为弱人工智能（狭义 AI），即设计用于执行特定任务；以及强人工智能（通用 AI），它将拥有全方位的类人认知能力。
2. The Turing Test and Measuring Intelligence | 图灵测试与智能的衡量
Proposed by Alan Turing in 1950, the Turing Test is an operational definition of intelligence. In the ‘imitation game’, a human interrogator communicates with both a human and a machine via text. If the interrogator cannot reliably distinguish the machine from the human, the machine is said to have passed the test and exhibited human-level linguistic intelligence. The test remains influential, though critics argue it only assesses conversation and not consciousness or true understanding.
图灵测试由艾伦·图灵于1950年提出，是智能的一种可操作定义。在“模仿游戏”中，一位人类询问者通过文本与另一个人和一台机器交谈。如果询问者无法可靠地区分机器和人类，就说这台机器通过了测试并展现出人类水平的语言智能。该测试至今仍具有影响力，但批评者认为它只评估对话能力，而非意识或真正的理解。
3. Rational Agents and the PEAS Model | 理性智能体与 PEAS 模型
An agent is anything that can perceive its environment through sensors and act upon that environment through actuators. A rational agent selects the action that is expected to maximise its performance measure, given its built-in knowledge and the percept sequence it has received. The PEAS framework (Performance, Environment, Actuators, Sensors) helps define the design task for an AI system.
智能体是指任何能够通过传感器感知环境并通过执行器在该环境中行动的事物。理性智能体会根据其内置知识和接收到的感知序列，选择预期能最大化性能度量的行动。PEAS框架（性能、环境、执行器、传感器）有助于定义 AI 系统的设计任务。
Consider an autonomous taxi as an example, with its PEAS description summarised in the table below.
以自动驾驶出租车为例，其 PEAS 描述总结于下表中。
| PEAS Component | Autonomous Taxi Example |
|---|---|
| Performance | Safety, minimise journey time, obey traffic laws, passenger comfort |
| Environment | Roads, other vehicles, pedestrians, traffic signals, weather conditions |
| Actuators | Steering, accelerator, brake, indicators, horn |
| Sensors | Cameras, LIDAR, GPS, speedometer, odometer, engine sensors |
| PEAS 组成部分 | 自动驾驶出租车示例 |
|---|---|
| 性能 | 安全、最短行程时间、遵守交通法规、乘客舒适度 |
| 环境 | 道路、其他车辆、行人、交通信号、天气条件 |
| 执行器 | 方向盘、油门、刹车、转向灯、喇叭 |
| 传感器 | 摄像头、激光雷达、GPS、速度计、里程表、发动机传感器 |
4. Problem Solving and Search Algorithms | 问题求解与搜索算法
Many AI tasks can be formulated as search problems, defined by an initial state, a goal test, actions that transition between states, and a path cost function. Uninformed (blind) search strategies have no domain knowledge beyond the problem definition.
许多 AI 任务可以被形式化为搜索问题，由初始状态、目标测试、状态间的转换动作以及路径成本函数定义。无信息（盲目）搜索策略除了问题定义外没有领域知识。
- Breadth-First Search (BFS) explores nodes level by level using a queue. It is complete and optimal if path cost is uniform, but uses significant memory.
- 广度优先搜索 (BFS) 使用队列逐层探索节点。若路径成本相同，它具备完备性和最优性，但内存开销大。
- Depth-First Search (DFS) explores a branch as far as possible before backtracking, using a stack (or recursion). It requires less memory than BFS but may not find the optimal solution and can get stuck in infinite paths without cycle checking.
- 深度优先搜索 (DFS) 在使用栈（或递归）时，尽可能深入地探索一个分支，然后回溯。它比 BFS 所需内存少，但可能找不到最优解，若无环路检查还会陷入无限路径中。
Informed (heuristic) search uses problem-specific knowledge to guide the search. Greedy Best-First Search expands the node with the lowest heuristic value h(n). It is not optimal. A* search combines path cost g(n) and heuristic h(n) to evaluate nodes using f(n) = g(n) + h(n). A* is optimal if the heuristic is admissible (never overestimates the true cost) and consistent.
有信息（启发式）搜索利用特定问题的知识来引导搜索。贪心最佳优先搜索扩展具有最低启发式值 h(n) 的节点。它不是最优的。A* 搜索结合路径成本 g(n) 和启发式 h(n)，使用 f(n) = g(n) + h(n) 评估节点。若启发式是可接受的（永不高估真实成本）且一致的，A* 即为最优。
Adversarial search applies in games. The minimax algorithm assumes both players play optimally: MAX tries to maximise the score, while MIN tries to minimise it. The algorithm constructs a game tree and backtracks utility values from the leaves. Alpha-beta pruning speeds up minimax by eliminating branches that cannot influence the final decision.
对抗搜索适用于博弈。极小极大算法假设双方都采取最佳策略：MAX 方力求最大化得分，MIN 方力求最小化得分。该算法构建博弈树并从叶子节点回溯效用值。Alpha-beta 剪枝通过剪除不会影响最终决策的分支来加速极小极大搜索。
5. Knowledge Representation and Reasoning | 知识表示与推理
To reason intelligently, an AI must represent knowledge in a formal language. Propositional logic uses atomic propositions and connectives: ¬ (not), ∧ (and), ∨ (or), → (implies), ↔ (if and only if). For example, ‘If it is raining then the ground is wet’ can be written as R → W. Truth tables determine the validity of formulae.
要智能地进行推理，AI 必须用形式化语言表示知识。命题逻辑使用原子命题和联结词：¬（非）、∧（与）、∨（或）、→（蕴含）、↔（当且仅当）。例如，“如果下雨，则地面湿”可写为 R → W。真值表用来确定公式的有效性。
First-order predicate logic extends propositional logic with objects, predicates, and quantifiers. For instance, ‘All students like AI’ can be written as ∀x Student(x) → Likes(x, AI). The universal quantifier ∀ means ‘for all’, and the existential quantifier ∃ means ‘there exists’. Resolution is a powerful inference rule used to derive new knowledge from a knowledge base.
一阶谓词逻辑通过引入对象、谓词和量词扩展了命题逻辑。例如，“所有学生都喜欢 AI”可以写成 ∀x Student(x) → Likes(x, AI)。全称量词 ∀ 意为“对所有”，存在量词 ∃ 意为“存在”。归结是一种强大的推理规则，用于从知识库中推导出新知识。
6. Expert Systems | 专家系统
An expert system emulates the decision-making ability of a human expert. It consists of a knowledge base (facts and rules), an inference engine that applies those rules, and a user interface. Rules are often expressed as IF-THEN structures. Forward chaining starts from known facts and applies rules to reach a conclusion; it is data-driven. Backward chaining starts from a goal and works backwards to find supporting evidence; it is goal-driven.
专家系统模拟人类专家的决策能力。它由知识库（事实和规则）、应用这些规则的推理引擎以及用户界面组成。规则通常表示为 IF-THEN 结构。正向链从已知事实出发并应用规则以得出结论；它是数据驱动的。反向链从目标出发，逆向寻找支撑证据；它是目标驱动的。
Expert systems offer advantages such as consistency, availability, and the ability to explain reasoning. However, they are expensive to build, difficult to maintain, and struggle with uncertainty unless fuzzy logic or certainty factors are incorporated.
专家系统的优点包括一致性、可用性和能够解释推理过程。然而，构建成本高昂，维护困难，并且除非引入模糊逻辑或置信度因子，否则难以处理不确定性。
7. Introduction to Machine Learning | 机器学习简介
Machine learning (ML) gives computers the ability to learn from data without being explicitly programmed. The three main paradigms are:
机器学习使计算机能够从数据中学习而无需明确编程。三种主要范式为：
- Supervised learning trains models on labelled data (input-output pairs). Common tasks include classification (predicting a category) and regression (predicting a continuous value).
- 监督学习使用带标签的数据（输入-输出对）训练模型。常见任务包括分类（预测类别）和回归（预测连续值）。
- Unsupervised learning finds hidden patterns in unlabelled data. Clustering groups similar data points, and dimensionality reduction simplifies data while retaining important features.
- 无监督学习从无标签数据中发现隐藏模式。聚类将相似的数据点分组，降维则在保留重要特征的同时简化数据。
- Reinforcement learning trains an agent through trial and error; the agent receives rewards or penalties and learns a policy to maximise cumulative reward.
- 强化学习通过试错来训练智能体；智能体获得奖励或惩罚，并学习最大化累积奖励的策略。
Key considerations in ML include bias-variance trade-off, overfitting, and the need for large, high-quality datasets.
机器学习中的关键问题包括偏差-方差权衡、过拟合，以及对大规模高质量数据集的需求。
8. Neural Networks and Deep Learning | 神经网络与深度学习
A neural network is a computational model inspired by the brain. The basic unit is the perceptron, which computes a weighted sum of its inputs, adds a bias, and passes the result through an activation function:
神经网络是一种受大脑启发的计算模型。基本单元是感知器，它计算输入的加权和，加上偏置，并将结果传递给激活函数：
output = f( ∑i wi xi + b )
where wi are weights, xi are inputs, b is the bias, and f is the activation function (e.g., step, sigmoid, ReLU). A multi-layer perceptron (MLP) stacks several layers of perceptrons, with hidden layers between input and output.
其中 wi 是权重，xi 是输入，b 是偏置，f 是激活函数（例如阶跃、Sigmoid、ReLU）。多层感知器 (MLP) 堆叠了多个感知器层，在输入和输出之间具有隐藏层。
Training a neural network uses backpropagation, which computes the gradient of the loss function with respect to each weight by applying the chain rule backwards through the network, enabling weights to be updated via gradient descent. Deep learning refers to neural networks with many hidden layers, enabling the learning of hierarchical
Published by TutorHao | A-Level 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