Artificial Intelligence (AI): How Machines Learn and Think | 人工智能：机器如何学习与思考

📚 Artificial Intelligence (AI): How Machines Learn and Think | 人工智能：机器如何学习与思考

Artificial intelligence (AI) is a branch of computer science that creates systems able to perform tasks that normally require human intelligence, such as recognising images, understanding speech, making decisions and translating languages. In Cambridge IGCSE Science, AI appears as an application of data, algorithms and automation, showing how scientists and engineers design machines that can learn from evidence. This article explains key AI ideas in a scientific way, using clear examples, and discusses both the benefits and the limitations of these systems.

人工智能（AI）是计算机科学的一个分支，它创建的系统能够执行通常需要人类智能的任务，例如识别图像、理解语音、做出决策和翻译语言。在剑桥 IGCSE 科学课程中，AI 作为数据、算法和自动化的应用出现，展示科学家和工程师如何设计能够从证据中学习的机器。本文以科学的方式解释 AI 的关键思想，使用清晰的例子，并讨论这些系统的优点与局限。

1. What Is Artificial Intelligence? | 什么是人工智能？

Artificial intelligence is the ability of a computer or a robot controlled by a computer to do tasks that are usually done by humans because they require human intelligence and judgement. Examples include identifying a plant from a photograph, predicting whether a patient has a disease, or controlling a self-driving car. AI systems are built from three main parts: input data, a model or algorithm, and an output decision or prediction.

人工智能是计算机或由计算机控制的机器人执行通常由人类完成的任务的能力，因为这些任务需要人类的智能和判断。例子包括从照片中识别植物、预测患者是否患病，或控制自动驾驶汽车。AI 系统由三个主要部分组成：输入数据、模型或算法，以及输出决策或预测。

In science, AI is often used to find patterns in large data sets that would be too complex for a human to analyse by hand. A climate scientist might use AI to spot trends in temperature records, while a biologist might use AI to sort thousands of microscope images.

在科学中，AI 常用于在大型数据集中寻找模式，这些模式对于人工分析来说过于复杂。气候科学家可能使用 AI 来发现温度记录中的趋势，而生物学家可能使用 AI 对数千张显微镜图像进行分类。

2. How AI Differs from Ordinary Software | AI 与普通软件有何不同？

Ordinary software follows a fixed set of rules written by a programmer. For example, a calculator always performs the same operation when you press the same buttons. AI software, however, can change its behaviour based on data. A machine learning model is not given exact rules for every situation; instead, it learns its own rules from examples.

普通软件遵循程序员编写的固定规则集。例如，当你按下相同的按钮时，计算器总是执行相同的操作。然而，AI 软件可以根据数据改变其行为。机器学习模型并没有为每种情况给出确切的规则；相反，它从示例中学习自己的规则。

This difference matters in science because many real-world problems, such as recognising handwritten chemical symbols or detecting a faint signal in noisy data, are difficult to describe with simple if-then rules. AI can learn the patterns automatically.

这种差异在科学中很重要，因为许多现实世界的问题，例如识别手写的化学符号或检测噪声数据中的微弱信号，很难用简单的 if-then 规则来描述。AI 可以自动学习这些模式。

3. Data: The Fuel of AI | 数据：AI 的燃料

AI models learn from data. Data can be numbers, images, sounds, text, or sensor readings. In a scientific experiment, data might include temperature, mass, pH, reaction time, or genetic sequences. The quality and quantity of data strongly affect how well an AI model performs.

AI 模型从数据中学习。数据可以是数字、图像、声音、文本或传感器读数。在科学实验中，数据可能包括温度、质量、pH、反应时间或基因序列。数据的质量和数量强烈影响 AI 模型的表现。

A useful way to think of data is as labelled examples. For a model that detects cancer cells, the training data might be thousands of microscope images, each labelled “healthy” or “cancerous”. The model learns to connect image features with the correct label.

理解数据的一个有用方式是将其视为带标签的示例。对于检测癌细胞的模型，训练数据可能是数千张显微镜图像，每张都标有 “健康” 或 “癌变”。模型学习将图像特征与正确的标签联系起来。

The table below shows common types of scientific data used in AI.

下表显示了 AI 中使用的常见科学数据类型。

Data type	Example in science
Numerical	Temperature readings over time
Image	Microscope slides of cells
Text	Research paper abstracts
Sensor	Acceleration from a motion detector

4. Algorithms and Models | 算法与模型

An algorithm is a step-by-step procedure for solving a problem. An AI model is the result of running a learning algorithm on data. The model contains numerical parameters, such as weights, that the algorithm adjusts during training to improve predictions.

算法是解决问题的逐步过程。AI 模型是在数据上运行学习算法的结果。模型包含数值参数，例如权重，算法在训练期间调整这些参数以改善预测。

A simple scientific example is linear regression, which finds the best straight line through data points. The model is the equation of the line: y = mx + c, where m is the gradient and c is the y-intercept. The learning algorithm adjusts m and c to minimise the difference between predicted and actual y values.

一个简单的科学例子是线性回归，它找到穿过数据点的最佳直线。模型是直线方程：y = mx + c，其中 m 是斜率，c 是 y 截距。学习算法调整 m 和 c，以最小化预测 y 值与实际 y 值之间的差异。

y = mx + c

5. Machine Learning and Training | 机器学习与训练

Machine learning is the process by which an AI model improves its performance by experience. Training means giving the model many examples and updating its parameters step by step. A common training method is called supervised learning, where every training example has a known correct answer.

机器学习是 AI 模型通过经验提高性能的过程。训练意味着给模型大量示例并逐步更新其参数。一种常见的训练方法称为监督学习，其中每个训练示例都有一个已知的正确答案。

The training process usually follows these steps: collect and clean data, split the data into a training set and a test set, train the model on the training set, and check accuracy on the test set. Splitting data helps scientists see whether the model can generalise to new examples instead of just memorising the training data.

训练过程通常遵循以下步骤：收集和清理数据，将数据分为训练集和测试集，在训练集上训练模型，并在测试集上检查准确率。划分数据有助于科学家了解模型是否能够推广到新示例，而不是仅仅记住训练数据。

Collect and clean data
Split into training and test sets
Train the model
Evaluate accuracy

这些步骤依次为：收集和清理数据，划分训练集和测试集，训练模型，评估准确率。

6. Neural Networks and Deep Learning | 神经网络与深度学习

A neural network is an AI model inspired by the brain. It is made of layers of connected units called neurons. Each connection has a weight. A neuron adds up the weighted inputs and passes the sum through an activation function to produce an output. This can be written as:

神经网络是一种受大脑启发的 AI 模型。它由称为神经元的连接单元层组成。每个连接都有一个权重。神经元将加权输入相加，并将总和通过激活函数以产生输出。这可以写成：

output = activation(w₁x₁ + w₂x₂ + b)

Here x₁ and x₂ are inputs, w₁ and w₂ are weights, and b is a bias term. The activation function decides whether the neuron ‘fires’. Deep learning simply means using many layers of neurons, allowing the model to learn more abstract features, such as edges in an image or sounds in speech.

这里 x₁ 和 x₂ 是输入，w₁ 和 w₂ 是权重，b 是偏置项。激活函数决定神经元是否 “激发”。深度学习只是指使用多层神经元，使模型能够学习更抽象的特征，例如图像中的边缘或语音中的声音。

7. Sensors and Real-World Inputs | 传感器与现实世界输入

Many AI applications in science depend on sensors that collect real-world data. Sensors measure physical quantities such as light, temperature, pressure, force, pH, or motion. The readings are converted into electrical signals and then into digital data that an AI model can process.

科学中许多 AI 应用依赖于收集现实世界数据的传感器。传感器测量物理量，如光、温度、压力、力、pH 或运动。读数被转换为电信号，然后转换为 AI 模型可以处理的数字数据。

For example, a weather station may use a temperature sensor, humidity sensor, and wind speed sensor. An AI model can combine these readings to predict the chance of rain. In a biology lab, a motion sensor and a camera can track the movement of a small organism for behaviour studies.

例如，气象站可能使用温度传感器、湿度传感器和风速传感器。AI 模型可以结合这些读数来预测下雨的可能性。在生物实验室中，运动传感器和摄像机可以追踪小生物的运动以进行行为研究。

8. Applications of AI in Science | AI 在科学中的应用

AI is used across many branches of science. In medicine, AI models analyse X-rays and MRI scans to help doctors detect diseases earlier. In chemistry, AI predicts the properties of new molecules and speeds up drug discovery. In environmental science, AI analyses satellite images to monitor deforestation, ocean pollution, and glacier melting.

AI 用于科学的许多分支。在医学中，AI 模型分析 X 射线和 MRI 扫描，帮助医生更早发现疾病。在化学中，AI 预测新分子的性质并加速药物发现。在环境科学中，AI 分析卫星图像以监测森林砍伐、海洋污染和冰川融化。

In physics and astronomy, AI helps process enormous amounts of data from particle accelerators and telescopes, identifying rare events such as a new exoplanet or a gravitational wave. These examples show that AI is not just a computer tool; it is becoming a partner in scientific discovery.

在物理学和天文学中，AI 帮助处理来自粒子加速器和望远镜的海量数据，识别罕见事件，如

Published by TutorHao | IGCSE Science Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）

Artificial Intelligence (AI): How Machines Learn and Think | 人工智能：机器如何学习与思考