📚 IB Computer Science: Artificial Intelligence Key Concepts | IB 计算机科学：人工智能 考点精讲
Artificial Intelligence (AI) is one of the most dynamic and examined topics in the IB Computer Science syllabus. This article distills all essential concepts, definitions, algorithms and ethical considerations you need for your HL Paper 2 or IA exploration. We will go through Turing tests, search strategies, expert systems, machine learning, neural networks, NLP, robotics and more, always linking theory to exam-style questions.
人工智能（AI）是 IB 计算机科学大纲中最活跃、最常考的专题之一。本文浓缩了 HL Paper 2 或 IA 探究中所需的所有核心概念、定义、算法和伦理考量。我们将依次梳理图灵测试、搜索策略、专家系统、机器学习、神经网络、自然语言处理、机器人学等内容，并始终把理论与考试题型结合起来。
1. Defining Artificial Intelligence | 人工智能的定义
In IB Computer Science, AI is defined as the theory and development of computer systems able to perform tasks normally requiring human intelligence. These tasks include visual perception, speech recognition, decision-making, and language translation. A key distinction is between ‘weak AI’ (systems designed for a specific task, like a chess engine) and ‘strong AI’ (systems that possess general intelligence comparable to a human). The syllabus emphasises that current AI is almost entirely weak AI, while strong AI remains a theoretical goal. Students should also understand the difference between symbolic AI (rule-based manipulation of symbols) and subsymbolic AI (e.g., neural networks processing numerical representations).
在 IB 计算机科学中，AI 被定义为能够执行通常需要人类智能的任务的计算机系统的理论与开发。这些任务包括视觉感知、语音识别、决策和语言翻译。一个关键区分是“弱人工智能”（为特定任务设计的系统，如国际象棋引擎）与“强人工智能”（具备可与人类媲美的通用智能的系统）。大纲强调目前几乎所有 AI 都是弱人工智能，强人工智能仍是理论目标。学生还应理解符号 AI（基于规则的符号操作）与亚符号 AI（如处理数值表示的神经网络）之间的区别。
2. Turing Test and Measuring Intelligence | 图灵测试与智能度量
Proposed by Alan Turing in 1950, the Turing test evaluates a machine’s ability to exhibit intelligent behaviour indistinguishable from a human. In the standard imitation game, a human interrogator asks questions via a text interface to both a human and a machine. If the interrogator cannot reliably tell which is the machine, the machine is said to have passed the test. IB exam questions often ask you to describe this setup and critique its limitations: the test measures linguistic competence rather than true understanding (Searle’s Chinese Room argument), and some modern chatbots can superficially pass constrained versions without genuine cognition. Alternatives like the Lovelace test (requiring creativity) or Winograd Schema (requiring common-sense reasoning) are also worth mentioning.
图灵测试由艾伦·图灵于 1950 年提出，用于评估机器表现出与人类无法区分的智能行为的能力。在标准的模仿游戏中，人类询问者通过文本界面向一个人和一台机器提问。如果询问者无法可靠地分辨哪个是机器，则该机器被称为通过了测试。IB 考试题常要求描述这一设置并批评其局限：该测试衡量的是语言能力而非真正的理解（塞尔的中文屋论证），而一些现代聊天机器人可以在没有真正认知的情况下表面通过受限版本。洛夫莱斯测试（要求创造力）或威诺格拉德模式（要求常识推理）等替代方案也值得提及。
3. Problem Solving and Search Algorithms | 问题解决与搜索算法
Many classical AI problems can be modelled as a search through a state space to find a goal state. The IB requires knowledge of blind (uninformed) and heuristic (informed) search algorithms. Breadth-first search (BFS) explores all nodes level by level, guaranteeing the shortest path in unweighted graphs but using exponential memory. Depth-first search (DFS) goes deep first and uses less memory but can get stuck in infinite loops. Heuristic searches like A* combine the actual cost so far g(n) and a heuristic estimate h(n) to the goal. The function f(n) = g(n) + h(n) is used; if h(n) is admissible (never overestimates), A* is optimal. Students should be able to trace these algorithms on simple graphs and explain the role of heuristics in reducing search time.
许多经典 AI 问题可以建模为在状态空间中搜索目标状态。IB 要求掌握盲目（无信息）搜索和启发式（有信息）搜索算法。广度优先搜索（BFS）逐层探索所有节点，保证在无权图中找到最短路径，但使用指数级内存。深度优先搜索（DFS）先深入探索，内存占用较少，但可能陷入无限循环。启发式搜索如 A* 结合了到目前为止的实际代价 g(n) 和到目标的启发式估计 h(n)。使用函数 f(n) = g(n) + h(n)；如果 h(n) 是可接受的（从不高估），则 A* 是最优的。学生应能追踪这些算法在简单图上的执行过程，并解释启发式在减少搜索时间中的作用。
4. Knowledge Representation and Reasoning | 知识表示与推理
For an AI to solve problems, knowledge must be stored in a form that enables reasoning. The IB covers semantic networks (graphs of concepts connected by labelled arcs, e.g., IS-A and HAS-A relationships), frames (data structures representing stereotypical situations with slots and default values), and production rules (IF-THEN statements used in expert systems). Reasoning techniques include forward chaining (data-driven: from known facts apply rules to derive new facts) and backward chaining (goal-driven: start from a hypothesis and see if the facts support it). You should be able to compare these: forward chaining can generate many irrelevant conclusions; backward chaining is more focused but requires a clear goal. Knowledge bases often face the frame problem (how to represent that most things stay the same when an action occurs).
为了让 AI 解决问题，知识必须以支持推理的形式存储。IB 涵盖语义网络（由带标签的弧连接概念的图，例如 IS-A 和 HAS-A 关系）、框架（表示典型情景及其槽和默认值的数据结构）和产生式规则（用于专家系统的 IF-THEN 语句）。推理技术包括前向链接（数据驱动：从已知事实出发，应用规则推导出新事实）和后向链接（目标驱动：从假设出发，检查事实是否支持）。你应能比较二者：前向链接可能产生许多无关结论；后向链接更聚焦，但要求明确的目标。知识库经常面临框架问题（如何表示当动作发生时，大多数事物保持不变）。
5. Expert Systems | 专家系统
An expert system is a computer program that emulates the decision-making ability of a human expert in a narrow domain (e.g., medical diagnosis, mineral prospecting). The IB structure includes: the knowledge base (facts and rules acquired from human experts via knowledge engineering), the inference engine (which applies rules to the facts using forward or backward chaining), the working memory (stores current facts during a consultation), and the explanation facility (which shows the reasoning chain, answering ‘how’ and ‘why’ questions). You should be able to draw the architecture and explain how the system reaches a conclusion. Limitations include difficulty in acquiring complete, consistent knowledge and the inability to learn from experience unless explicitly reprogrammed.
专家系统是一种在狭窄领域（如医疗诊断、矿产勘探）模拟人类专家决策能力的计算机程序。IB 要求掌握其结构：知识库（通过知识工程从人类专家那里获取的事实和规则）、推理引擎（使用前向或后向链接将规则应用于事实）、工作内存（在咨询期间存储当前事实）和解释设施（显示推理链条，回答“如何”和“为什么”问题）。你应能画出架构图并解释系统如何得出结论。局限性包括难以获取完整一致的知识，以及除非明确重新编程否则无法从经验中学习。
6. Machine Learning Fundamentals | 机器学习基础
Machine learning (ML) is the study of algorithms that improve automatically through experience. The syllabus categorises ML into supervised learning (labelled training data, e.g., regression and classification with decision trees, k-nearest neighbours), unsupervised learning (unlabelled data, finding hidden patterns like clustering via k-means), and reinforcement learning (an agent learns to act in an environment to maximise cumulative reward). Students must understand the importance of training data quality, overfitting (model learns noise instead of genuine patterns) and underfitting (model is too simple to capture the underlying trend). Evaluation metrics like precision, recall, and F1-score may appear in Paper 2 scenarios. A common exam task is to describe the steps in building a classifier: collect data, preprocess, split into training/test sets, choose a model, train, evaluate, and adjust hyperparameters.
机器学习（ML）是研究通过经验自动改进的算法的领域。大纲将 ML 分为监督学习（有标签的训练数据，例如回归和使用决策树、k 近邻的分类）、无监督学习（无标签数据，通过 k-means 等聚类发现隐藏模式）和强化学习（智能体学习在环境中行动以最大化累积奖励）。学生必须理解训练数据质量的重要性、过拟合（模型学习噪声而非真实模式）和欠拟合（模型过于简单，无法捕捉潜在趋势）。评估指标如精确率、召回率和 F1 分数可能出现在 Paper 2 场景中。常见的考题是描述构建分类器的步骤：收集数据、预处理、分割训练/测试集、选择模型、训练、评估和调整超参数。
7. Neural Networks and Deep Learning | 神经网络与深度学习
The IB introduces artificial neural networks (ANNs) as a model inspired by biological neurons. The basic unit is the perceptron, which takes weighted inputs, sums them, applies an activation function (e.g., sigmoid, ReLU) and outputs a value. A network consists of an input layer, one or more hidden layers, and an output layer. Learning occurs by adjusting the weights through backpropagation: after forward propagation, the error between predicted and actual output is computed, and the gradient of this error with respect to each weight is propagated backwards to update weights via gradient descent. Key terms include learning rate, epoch, batch size, and loss function. While detailed calculus is not required, you should explain the process in logical steps and understand that deep learning refers to networks with many hidden layers, which can learn hierarchical features from raw data (e.g., edges to shapes to objects in image recognition).
IB 将人工神经网络（ANN）作为一种受生物神经元启发的模型来介绍。基本单元是感知机，它接收加权输入，对其求和，应用激活函数（如 sigmoid、ReLU）并输出一个值。网络由输入层、一个或多个隐藏层和输出层组成。学习是通过反向传播调整权重来实现的：在前向传播后，计算预测输出与实际输出之间的误差，然后将该误差相对于每个权重的梯度反向传播，通过梯度下降更新权重。关键术语包括学习率、epoch、批次大小和损失函数。虽然不要求详细的微积分，但你应能按逻辑步骤解释这一过程，并理解深度学习指的是具有多个隐藏层的网络，能够从原始数据中学习层次化特征（例如，图像识别中从边缘到形状再到物体）。
8. Natural Language Processing | 自然语言处理
Natural Language Processing (NLP) enables computers to understand, interpret and generate human language. IB topics include parsing (breaking down a sentence into syntactic components, often using context-free grammars or dependency trees), sentiment analysis (determining the emotional tone of a text), and machine translation. Early systems used rule-based translation, but modern approaches use statistical and neural machine translation (e.g., sequence-to-sequence models with attention). You should discuss the challenges: ambiguity (lexical, syntactic, referential), idioms, context and cultural nuances. The Turing test’s connection to NLP is strong, as passing the test requires fluent language use. Exam questions may ask you to explain how a chatbot processes an input sentence, from tokenisation to response generation.
自然语言处理（NLP）使计算机能够理解、解释和生成人类语言。IB 涵盖的主题包括句法分析（通常使用上下文无关文法或依存树将句子分解为句法成分）、情感分析（确定文本的情感倾向）和机器翻译。早期系统使用基于规则的翻译，但现代方法使用统计和神经机器翻译（例如带有注意力机制的序列到序列模型）。你应讨论其中的挑战：歧义（词汇、句法、指代）、习语、语境和文化细微差别。图灵测试与 NLP 的联系很强，因为通过测试需要流利的语言运用。考试问题可能要求你解释聊天机器人如何处理一个输入句子，从分词到响应生成。
9. Computer Vision | 计算机视觉
Computer vision focuses on enabling machines to interpret visual data from the world. Key processes include image acquisition, preprocessing (noise removal, normalisation), feature extraction (edges, corners, textures using filters like Sobel or methods like SIFT), and object recognition (classifying the object present, often using convolutional neural networks, CNNs). CNNs use layers of convolution, pooling and fully connected layers to automatically learn features. Applications include facial recognition, autonomous driving and medical imaging. The IB expects you to relate this to machine learning: for instance, a CNN is trained on thousands of labelled images to classify new ones. You should also mention ethical concerns regarding surveillance and bias in facial recognition.
计算机视觉专注于使机器能够解读来自世界的视觉数据。关键过程包括图像采集、预处理（去噪、归一化）、特征提取（使用 Sobel 滤波器或 SIFT 等方法提取边缘、角点、纹理）和物体识别（对存在的物体进行分类，常使用卷积神经网络 CNN）。CNN 使用卷积层、池化层和全连接层来自动学习特征。应用包括人脸识别、自动驾驶和医学成像。IB 要求你将其与机器学习联系起来：例如，CNN 在数千张带标签的图像上训练，以对新的图像进行分类。你还应提及关于监控和人脸识别中偏见的伦理关切。
10. Robotics | 机器人学
In IB Computer Science, robotics integrates sensors, actuators and control systems to create machines that can interact physically with the world. The classic sense-plan-act cycle is central: the robot senses its environment (using cameras, lidar, tactile sensors), plans actions based on its goals and world model, and then acts via motors or manipulators. AI is applied in path planning (e.g., A* on a grid map), simultaneous localisation and mapping (SLAM), and computer vision for object manipulation. Types of robots include manipulator arms, mobile robots, and humanoid robots. Social and ethical issues include autonomous weapons, job replacement in manufacturing, and care robots for the elderly. The IB often asks you to discuss a specific robotic scenario and identify inputs, processes and outputs.
在 IB 计算机科学中，机器人学集成了传感器、执行器和控制系统，以创建能与物理世界互动的机器。经典的感知-规划-行动循环是核心：机器人感知其环境（使用摄像头、激光雷达、触觉传感器），根据其目标和世界模型规划行动，然后通过电机或操纵器执行。AI 被应用于路径规划（例如在网格地图上使用 A*）、同步定位与地图构建（SLAM）以及用于物体操作的计算机视觉。机器人类型包括机械臂、移动机器人和人形机器人。社会与伦理问题包括自主武器、制造业的就业替代以及老年人护理机器人。IB 常要求你讨论一个特定的机器人场景，并识别输入、处理和输出。
11. Ethical and Social Implications of AI | AI 的伦理与社会影响
AI technology raises profound ethical questions that feature prominently in IB Paper 2 and the IA. Bias in AI systems can arise from skewed training data, leading to unfair outcomes in hiring, policing, or credit scoring. Privacy is eroded by pervasive data collection (smart speakers, facial recognition in public). Accountability is blurry when an autonomous vehicle causes harm-who is responsible? Employment is reshaped as AI automates both manual and cognitive tasks, creating new jobs while displacing others. The IB expects you to discuss these issues using specific examples, analyse stakeholders (users, developers, regulators, society), and suggest mitigation strategies such as algorithmic transparency, bias audits, and regulation. Avoiding generic statements is key; always tie ethical points to the AI technique being used.
AI 技术引发了深刻的伦理问题，这在 IB Paper 2 和 IA 中占有突出位置。AI 系统中的偏见可能源于训练数据的偏差，导致在招聘、警务或信用评分中出现不公平结果。无处不在的数据收集（智能音箱、公共场所的人脸识别）侵蚀了隐私。当自动驾驶车辆造成伤害时，问责变得模糊——谁该负责？随着 AI 自动执行体力劳动和认知任务，就业被重塑，在创造新岗位的同时取代了其他岗位。IB 要求你使用具体例子讨论这些问题，分析利益相关者（用户、开发者、监管机构、社会），并提出缓解策略，如算法透明度、偏见审计和监管。避免泛泛而谈至关重要；始终将伦理观点与所涉的 AI 技术联系起来。
12. Exam Tips and the Future of AI | 考试技巧与 AI 的未来
To excel in IB Computer Science AI questions, practice structuring your answers with clear definitions, diagrams where applicable (e.g., expert system architecture, neural network layers), and concrete examples. When asked to evaluate, always present both advantages and limitations. For the IA, if your project involves AI, document your data source, preprocessing steps, algorithm choice and why, testing methodology, and ethical considerations. Looking forward, AI research moves toward general AI, explainable AI (XAI), and multi-modal systems. While these are beyond the core syllabus, referencing them briefly when discussing trends can demonstrate broader understanding. Keep your answers concise, directly linked to the command terms (describe, explain, evaluate, etc.) from the markscheme.
要在 IB 计算机科学 AI 试题中脱颖而出，练习以清晰的定义、适用时配以图示（如专家系统架构、神经网络层）和具体例子来组织答案。当被要求评估时，务必同时展现优点和局限。对于 IA，如果你的项目涉及 AI，请记录数据来源、预处理步骤、算法选择及其理由、测试方法和伦理考量。展望未来，AI 研究正朝着通用 AI、可解释 AI（XAI）和多模态系统发展。虽然这些超出核心大纲，但在讨论趋势时简要提及可以展示更广的理解。保持答案简洁，直接对应评分方案中的指令词（描述、解释、评估等）。
Published by TutorHao | IB 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