6.3 · Artificial intelligence
6.3 · 人工智能
Artificial intelligence
人工智能
What AI is and its characteristics, then the only two AI systems the syllabus covers: expert systems (their four parts: knowledge base, rule base, inference engine, user interface) and machine learning.
什么是人工智能及其特征,然后是考纲仅涵盖的两种人工智能系统:专家系统(包含知识库、规则库、推理引擎和界面)以及机器学习。
Explainer · 6.3.1
讲解 · 6.3.1
What artificial intelligence is
什么是人工智能
Artificial intelligence (AI) is the branch of computer science that deals with getting a computer to simulate intelligent behaviour.
人工智能(AI)是计算机科学的一个分支,研究让计算机模拟智能行为。
The word to hold onto is simulate. The computer is not conscious and does not really understand; it produces behaviour that looks intelligent, such as answering a question, spotting a pattern or making a decision.
要抓住的词是模拟。计算机没有意识,也并非真正理解;它产生的是看上去智能的行为,例如回答问题、发现规律或做出决策。
The 0478 syllabus keeps AI small. You need what AI is, its characteristics, and two AI systems: expert systems and machine learning. Neural networks, deep learning and "narrow / general / strong AI" are not in this syllabus.
0478 考纲把人工智能控制在很小的范围。你需要掌握:什么是 AI、它的特征,以及两种人工智能系统:专家系统和机器学习。神经网络、深度学习以及"弱人工智能 / 通用人工智能 / 强人工智能"不在本考纲内。
Explainer · 6.3.2
讲解 · 6.3.2
The characteristics of AI
人工智能的特征
A question that asks for the characteristics of AI wants these, not a list of examples or types.
问人工智能特征的题目要的是下面这些,不是一串例子或类别。
| Characteristic | 特征 | What it means | 含义 |
|---|---|---|---|
| A body of facts, plus rules for what to do with those facts | 一批数据,以及使用这些数据的规则 | the system holds facts about its subject and rules that say what to do with them | 系统保存关于其主题的事实,以及规定如何处理这些事实的规则 |
| The ability to reason | 推理能力 | it can draw a conclusion from the data and rules, rather than just looking a fact up | 它能根据数据和规则得出结论,而不只是查一个事实 |
| In some cases, an ability to learn and adjust over time | 它可以包含学习和适应的能力 | some AI changes its own rules or data from experience, so its results improve over time (this one is optional, not every AI does it) | 有些人工智能会根据经验改变自己的规则或数据,使结果随时间改进(这一点是可选的,并非每个 AI 都会) |
Markers report that "characteristics" answers describe scenarios where AI is used, or types of AI. Neither answers the question. Give the three points above.
考官报告指出,"特征"类作答描述的是使用人工智能的场景,或人工智能的类型。这两者都没回答问题。要给出上面这三点。
Try it yourself
动手试试
Characteristic or not?
是特征还是不是?
Tap each statement to tag it as a characteristic of AI or not, then check.
点击每条陈述,把它标记为人工智能的特征还是不是,然后检查。
Explainer · 6.3.3
讲解 · 6.3.3
The two AI systems
两种人工智能系统
| System | 系统 | What it is | 是什么 | Fits a task where | 适用于这样的任务 |
|---|---|---|---|---|---|
| Expert system | 专家系统 | a computer that reaches a conclusion the way a human expert would, using stored facts and rules; it asks questions and works towards an answer | 一台像人类专家那样得出结论的计算机,它使用存储的事实和规则;它提出问题,朝着一个答案推进 | the knowledge can be written as fixed rules: diagnosing a fault or an illness, identifying a species, tax calculations | 知识可以写成固定规则:诊断故障或疾病、鉴定物种、税务计算 |
| Machine learning | 机器学习 | a program that changes its own processes and/or data by itself, from the examples it is given, so it improves without being reprogrammed | 一个能根据所给示例自动调整自己的处理方式和/或数据的程序,于是它无需重新编程就能改进 | the rules are hard to write out but there are many examples to learn from: sorting spam, recommending products, recognising speech | 规则难以写清,但有大量可供学习的示例:过滤垃圾邮件、推荐商品、识别语音 |
Try it yourself
动手试试
Expert system or machine learning?
专家系统还是机器学习?
Tap each task to tag it as one for an expert system or one for machine learning, then check.
点击每项任务,把它标记为适合专家系统还是机器学习的,然后检查。
Explainer · 6.3.3
讲解 · 6.3.3
Inside an expert system: four parts
专家系统的内部:四个部分
Markers report that candidates can name the parts but cannot describe their role. The role is what earns the marks. Use the exact terms.
考官报告指出,考生能说出这几个部分的名称,却说不出它们的作用。作用才是得分点。要用准确的术语。
| Part | 部分 | Its role | 它的作用 |
|---|---|---|---|
| Knowledge base | 知识库 | stores the facts about the subject, gathered from human experts and written sources | 存储关于该主题的事实,这些事实来自人类专家和文献资料 |
| Rule base | 规则库 | stores the rules that link the facts, written as IF…THEN statements | 存储把这些事实联系起来的规则,写成"如果…那么"的语句 |
| Inference engine | 推理引擎 | the processing part: it takes the user's answers, searches the knowledge base, applies the rules from the rule base, and works out the next question or the final conclusion | 处理部分:它接收用户的回答,搜索知识库,应用规则库中的规则,推算出下一个问题或最终结论 |
| Interface | 界面 | how the user and the system communicate: it presents the questions (often yes/no) and shows the conclusion and its probability | 用户与系统沟通的方式:它呈现问题(通常是是/否),并显示结论及其概率 |
Use the full, correct terms. "Rule data" does not earn the mark for rule base, and "interface engine" is not the inference engine.
要用完整、正确的术语。"规则数据"拿不到规则库的分,"界面引擎"也不是推理引擎。
Try it yourself
动手试试
Explore each part
逐一了解每个部分
Click each part of an expert system to see its role.
点击专家系统的每个部分,看看它的作用。
Worked example · 6.3.3
范例 · 6.3.3
How an expert system reaches an answer
专家系统如何得出答案
Mia's tomato plants are not doing well and she uses an expert system for gardeners. Markers want the parts described working together to reach the answer, not just listed.
Mia 的番茄植株长得不好,她使用一个面向园丁的专家系统。考官要的是把这几个部分描述成协同工作以得出答案,而不只是罗列出来。
The interface asks Mia a yes/no question. Her answer goes to the inference engine, which searches the knowledge base for matching facts and applies the rule base. Based on her answer it picks the next question. When the rules point to one cause, the interface shows the conclusion and how likely it is: "Most likely: not enough water (about 80%)."
界面向 Mia 提一个是/否问题。她的回答送到推理引擎,推理引擎在知识库中搜索匹配的事实,并应用规则库。根据她的回答,它选出下一个问题。当规则指向某一个原因时,界面显示结论及其可能性:"最可能:缺水(约 80%)。"
Try it yourself
动手试试
Follow a diagnosis
跟踪一次诊断
Step through the expert system helping Mia. Watch which part is doing the work at each stage.
逐步走一遍帮助 Mia 的这个专家系统。注意每个阶段是哪个部分在干活。
Explainer · 6.3.3
讲解 · 6.3.3
Machine learning
机器学习
In machine learning, a program is able to change its own processes and/or data by itself, learning from examples, so that its results get better over time without a programmer editing the code.
机器学习是指:一个程序能自动调整自己的处理方式和/或数据,从示例中学习,于是它的结果随时间变好,而无需程序员改动代码。
An ordinary program does exactly what it was written to do, every time. A machine learning program is trained on examples first, then keeps adjusting from new examples: a spam filter that is told "this was spam" changes its own rules so it catches more next time.
普通程序每次都严格按写好的方式运行。机器学习程序先用示例训练,然后不断根据新示例调整:一个被告知"这是垃圾邮件"的垃圾邮件过滤器会改变自己的规则,使它下次能拦下更多。
Markers report "it can now learn" with no detail. Say what it learns from (examples / past results) and what changes (its own rules or data), tied to the scenario in the question.
考官报告看到"它现在能学习了"这种没有细节的作答。要说清它从什么学(示例 / 过去的结果),以及什么发生了改变(它自己的规则或数据),并结合题目里的情境。
Try it yourself
动手试试
Fixed program or machine learning?
固定程序还是机器学习?
Tap each description to tag it as an ordinary fixed program or one using machine learning, then check.
点击每条描述,把它标记为普通的固定程序还是使用了机器学习的,然后检查。
Check your understanding
自测
Which of these is not a characteristic of AI?
以下哪一项不是人工智能的特征?
Check your understanding
自测
What does the inference engine do in an expert system?
在专家系统中,推理引擎的作用是什么?
Check your understanding
自测
What makes a program a machine learning program?
是什么让一个程序成为机器学习程序?
Common mistakes
常见错误
6.3.1 to 6.3.3
考官针对 6.3.1 至 6.3.3 报告了以下模式
RecurringOn expert systems, candidates name the parts but cannot describe their role, or list the parts as a generic answer without applying them to the scenario. Markers want the parts described working together to reach the answer for the case in the question.
反复出现关于专家系统,考生能说出各部分的名称,却说不出它们的作用,或者把各部分当成通用答案罗列,没有结合情境。考官要的是把这些部分描述成协同工作、为题目里的具体情况得出答案。
Imprecise terms. Markers do not accept "rule data" for rule base, or "interface engine" for inference engine. Learn the four terms exactly: knowledge base, rule base, inference engine, interface.
术语不准确。考官不接受用"规则数据"代替规则库,也不接受用"界面引擎"代替推理引擎。要把这四个术语记准:知识库、规则库、推理引擎、界面。
"Characteristics of AI" answered with examples or types. Describing smart speakers, or "narrow / general / strong AI", does not answer the question. Give the three: facts plus rules for using them, reasoning to a conclusion, and (sometimes) learning and adjusting.
"人工智能的特征"用例子或类别来回答。描述智能音箱、或"弱 / 通用 / 强人工智能",都没回答问题。要给出:一批数据和规则、推理能力,以及可选的学习和适应能力。
Machine learning described as just "it can learn". Say what it learns from and what it changes: it adapts its own rules or data from the examples it is given, tied to the scenario. And do not bring in neural networks or deep learning: they are outside 0478.
把机器学习只描述成"它能学习"。要说清它从什么学、改变了什么:它根据所给示例调整自己的规则或数据,并结合情境。另外不要扯到神经网络或深度学习:它们不在 0478 范围内。
Names, no roles: "An expert system is built from four parts: knowledge base, rule base, inference engine, user interface."
只有名称,没有作用:"专家系统有知识库、规则库、推理引擎和界面。"
Right: "The interface asks the user a yes/no question. The inference engine takes the answer, searches the knowledge base of facts, applies the IF...THEN rules from the rule base, and uses the result to choose the next question. When the rules point to one cause, the interface shows that conclusion and how likely it is."
正确:"界面向用户提一个是/否问题。推理引擎接收回答,搜索由事实构成的知识库,应用规则库中的"如果...那么"规则,并用结果来选出下一个问题。当规则指向某一个原因时,界面显示该结论及其可能性。"
No detail: "The robot uses machine learning so it can now learn and get better."
没有细节:"机器人使用机器学习,所以它现在能学习、能变得更好。"
Right: "Each time the robot meets an obstacle it stores what the obstacle looked like and which move got it past. It adds this to its own data, so next time it meets a similar obstacle it chooses the successful move straight away."
正确:"机器人每次遇到障碍,都会记下障碍的样子以及哪个动作让它通过了。它把这个加进自己的数据里,于是下次遇到类似障碍时,它会直接选那个成功的动作。"
Go deeper
深入练习
Walk an expert system through a scenario
让专家系统走一遍某个情境
Pick a use for an expert system (a fault finder, an illness checker, a species identifier). Write out how the interface, inference engine, knowledge base and rule base work together, step by step, to reach one specific conclusion. Then write two sentences on how a machine learning system would improve at the same task over time. Check against 6.3.3.
给专家系统挑一个用途(故障查找器、病症检查器、物种鉴定器)。逐步写出界面、推理引擎、知识库和规则库如何协同工作,得出某一个具体结论。然后用两句话写出机器学习系统在同一任务上如何随时间改进。对照 6.3.3 核对。
Revision questions
复习题
Past-paper style. Write an answer, then reveal the marking points and mark yourself.
采用历年真题风格。写下答案,然后展开评分要点给自己打分。
[1]State what artificial intelligence is.
[1]陈述人工智能的含义。
[3]Describe the features that make a system count as AI.
[3]描述人工智能的主要特征。
Marking points
评分要点
- Any three: it holds data / facts and a set of rules for using them; it can use the rules to reason towards a conclusion or make a decision; it can (in some systems) learn and adapt from experience or new data.
- 任选三点:它拥有数据 / 事实以及使用它们的一组规则;它能用规则推理得出结论或做出决定;它(在某些系统中)能从经验或新数据中学习和适应。
[4]Describe the role of each of the four parts of an expert system.
[4]描述专家系统四个部分各自的作用。
Marking points
评分要点
- Knowledge base: stores the facts about the subject.
- Rule base: stores the IF...THEN rules for using the facts.
- Inference engine: applies the rules to the facts and the user's answers to reason towards a conclusion, and chooses the next question.
- User interface: asks the user questions and presents the conclusion.
- 知识库:存储关于该主题的事实。
- 规则库:存储使用这些事实的 IF...THEN 规则。
- 推理引擎:把规则应用于事实和用户的回答,朝结论进行推理,并选择下一个问题。
- 用户界面:向用户提问,并给出结论。
[5]A garage uses an expert system to help mechanics find engine faults. Describe how the parts of the expert system work together to reach a diagnosis.
[5]一家修车厂用专家系统帮助技工查找发动机故障。描述该专家系统的各个部分如何协同工作,得出诊断结果。
Marking points
评分要点
- The interface asks the mechanic questions about the engine's symptoms.
- Each answer is passed to the inference engine.
- The inference engine searches the knowledge base of facts about engine faults and applies the IF...THEN rules from the rule base.
- Based on the answers it selects the next question to narrow down the cause.
- When the rules lead to a single fault, the interface presents that diagnosis (often with a recommended repair).
- 界面向技工询问发动机症状的问题。
- 每个回答被传给推理引擎。
- 推理引擎搜索关于发动机故障的事实知识库,并应用规则库中的 IF...THEN 规则。
- 根据回答,它选择下一个问题以缩小原因范围。
- 当规则指向唯一的故障时,界面给出该诊断(通常附带建议的维修)。
[2]Explain what machine learning is.
[2]解释机器学习的含义。
Marking points
评分要点
- Machine learning is when a program improves its own performance over time by learning from examples or past results.
- It adjusts its own rules or data, without a programmer changing the code.
- 机器学习是指程序随时间推移,通过从示例或过去的结果中学习来改进自身表现。
- 它调整自己的规则或数据,而无需程序员修改代码。
Flashcards
记忆卡片
Active recall for this subtopic’s key terms.
针对本子主题关键术语的主动回忆练习。
Keep this subtopic’s practice with you
把这个子主题的练习保存下来
Free account: unlock every revision question and flashcard for 6.3, and every other IGCSE 0478 statement.
免费账户:解锁 6.3 以及所有其他 IGCSE 0478 考纲陈述的全部复习题和记忆卡片。