Appendix A. Glossary
These terms are grouped not in alphabetical order, but roughly in the order they first appear in the book. The chapter in parentheses indicates where the term was first covered in detail.
| Term | Meaning |
|---|---|
| AI (Artificial Intelligence) | A general term for technologies that enable computers to understand, judge, and create like humans. (Chapter 1) |
| Generative AI | AI that mimics the patterns and characteristics of large amounts of data to create new outputs such as text, images, sounds, and videos. It is defined in Article 2, Item 5 of the AI Basic Act. (Chapter 1) |
| Chatbot | A program that asks and answers questions in writing or speech. A portmanteau of chat (conversation) and robot. (Chapter 1) |
| Conversational AI | AI that works through conversations exchanged with humans. In this book, it was used with the same meaning as chatbot. (Chapter 1) |
| Chat window | The screen where you write questions and read answers. It is safe to think of it like a letter sent to a company. (Chapters 1 and 4) |
| Input box | The field at the bottom of the chat window where you write questions. (Chapter 2) |
| Prompt | A question or request sent to an AI. In this book, it was referred to as a "question" or "request." (Chapter 1) |
| Answer (Response) | The text sent back by the AI. Even if it sounds smooth and confident, there is no guarantee that it is correct. (Chapter 1) |
| Training | The process in which an AI learns the flow of language by reading large amounts of text. It also refers to using user conversations for AI improvement. (Chapters 1 and 4) |
| Next-word continuation | The method by which generative AI creates answers by stringing together, piece by piece, natural-sounding words to follow a question. The criterion is naturalness, not facts. (Chapter 1) |
| Training cutoff (Knowledge cutoff date) | The reference point in time for the data learned by the AI. It may not know or may state incorrectly any timetables, fares, or systems changed after that time. (Chapter 1) |
| Hallucination (Illusion) | A phenomenon where an AI states things that are not true as if they were true. It may fabricate non-existent stores, books, or sources. (Chapter 1) |
| Search | A feature that finds a list of websites containing information. It is suitable for checking today's facts. (Chapter 1) |
| Nurijip | The Korean word meaning website. (Chapters 1 and 3) |
| Source | Where information came from. Go directly to the official website, find the same document, and verify four things: can it be found, who wrote it, is the content the same, and when was it written. (Chapter 3) |
| Three-column division | A method of dividing tasks into "leave it to AI / AI assists and I verify / I do it myself." (Chapter 1) |
| Follow-up questioning | Asking again in the same chat window following previous conversations. The AI reviews the previous content together. (Chapter 2) |
| Refining the question | Pinpointing what is lacking after seeing the first answer and asking the AI to refine it again. You should also tell it to keep the good parts. (Chapter 2) |
| New chat | A newly opened chat window. Because whether the previous conversation carries over depends on the app and settings (such as memory features), write down necessary conditions again. Open one when the topic changes or the conversation gets tangled. (Chapter 2) |
| Five-box question chart | What, why, to whom, how much, and in what format. Five things to include in a good question. (Chapter 2) |
| Situation (Context) | The purpose, occasion, conditions, already decided matters, and things you want to avoid in a question. (Chapter 2) |
| Audience | The person who will read or receive the output. It serves as the standard for deciding tone and vocabulary. (Chapter 2) |
| Format | The shape of the output. Lists, tables, numbered steps, short lines without tables, etc. Tailor it to where it will be used. (Chapter 2) |
| Leaving blanks | A method of having the AI leave parts that are easy for AI to fabricate, such as dates, locations, and store names, blank like [Date], and filling in the facts yourself. (Chapter 2) |
| Requesting questions back | A method of asking the AI to ask questions first when you do not know what information to provide. (Chapter 2) |
| Eight items to underline | Numbers, dates and times, people's names, names of stores, institutions, and systems, locations and addresses, telephone numbers and website addresses, qualifications and rules, and rationales. Areas to verify in an AI answer. (Chapter 3) |
| Official website | A website operated by the institution directly responsible for that information. The first priority for verification. (Chapter 3) |
| Main telephone number | An institution's telephone number listed on its official website. Call this number instead of the number written in an AI answer or text message. (Chapter 3) |
| Publication date and modification date | The date a post was published or modified. A clue to judge whether information is still accurate today. (Chapter 3) |
| Enforcement date | The date on which a law or system actually begins to be applied. The National Law Information Center displays it below the name of the law. (Chapter 3) |
| Cross-checking | Checking whether the content matches across two or more places. A single official guide is better than multiple blogs. (Chapter 3) |
| National Law Information Center | A website operated by the Ministry of Government Legislation where you can view original texts of legal provisions and enforcement dates. (Chapter 3) |
| Easy-to-Understand Everyday Law Information | A website operated by the Ministry of Government Legislation that explains everyday legal issues in plain language. (Chapter 3) |
| National Health Information Portal | A website operated by the Korea Disease Control and Prevention Agency that contains general information on diseases and health. (Chapter 3) |
| Privacy | Information relating to a living individual that identifies that individual. It also includes information that can be easily combined with other information to identify an individual (Article 2, Item 1 of the Personal Information Protection Act (PIPA)). (Chapter 4) |
| Data subject | A person who can be identified by the processed information, that is, the owner of that information. They have rights such as requesting access, correction or deletion, and suspension of processing (Article 4). (Chapter 4) |
| Personal information controller | A public institution, legal entity, organization, individual, etc., that processes personal information in order to operate personal information files for work purposes (Article 2, Item 5). (Chapter 4) |
| Sensitive information | Information such as health, ideology, and beliefs that can significantly infringe upon privacy, the processing of which is separately restricted by law (Article 23). (Chapter 4) |
| Restriction on resident registration number processing | A provision allowing personal information controllers to process resident registration numbers only when there is specific legal authority in statutes, etc. (Article 24-2). (Chapter 4) |
| Verification code | A multi-digit number received via text message or other means for identity verification. Never disclose it to anyone. (Chapter 4) |
| Server | A computer operated by a service company. Text and photos written in the chat window go here. (Chapter 4) |
| Chat history | A list of past conversations. Even if deleted, it may not immediately disappear from everywhere. (Chapter 4) |
| Training settings | Settings that determine whether to use your conversations for AI Model Improvement. Names and locations differ across services. (Chapter 4) |
| Model Improvement | The act of making AI better using user conversations, etc. (Chapter 4) |
| Human review | A human reading and reviewing conversations for service improvement, etc. (Chapter 4) |
| Temporary chat | A chat feature that does not remain in the chat history. It still passes through servers and can be stored briefly. (Chapter 4) |
| Memory | A feature in some services that remembers conversation content and uses it in subsequent conversations. (Chapter 4) |
| Traffic light classification | A method of classifying information into Red (do not enter), Yellow (mask before entering), and Green (safe to enter). (Chapter 4) |
| Four masking methods | Omission, substitution, generalization, and deletion. Methods for erasing personal information before asking questions. (Chapter 4) |
| Shared device | A device used by multiple people together, such as computers in libraries or community centers, or a shared family tablet. When done using it, click log out and verify. (Chapter 4) |
| Log out | Exiting your account. This is different from closing the window. (Chapter 4) |
| Deepfake | Fake photos, videos, or voices created by mimicking a person's face or voice with AI. (Chapter 5) |
| Deepvoice | A type of deepfake that mimics only the voice. (Chapter 5) |
| Phishing | Fraud that deceives people to extract money or personal information. When conducted over the phone, it is called voice phishing. (Chapter 5) |
| Smishing | Phishing using text messages. It leads victims to fake websites through links in text messages or causes them to install malicious apps. (Chapter 5) |
| Messenger family impersonation | Fraud where someone pretends to be a family member on a messaging app and demands photos of credit cards, verification codes, or wire transfers. Verify by calling the saved phone number. (Chapter 5) |
| Malicious app | An app that causes harm, such as stealing information or intercepting phone calls, once installed on a mobile phone. (Chapter 5) |
| Caller ID spoofing | Manipulating an incoming call so that a different number is displayed on the screen. Do not believe a call is genuine based solely on the number on the screen. (Chapter 5) |
| Five signs of fraudulent contact | Urgency, secrecy, demands for money or financial information, links or app installations, and unfamiliar numbers or messaging apps. (Chapter 5) |
| Family verification question | A question decided in advance by family members meeting in person that only family members can answer. (Chapter 5) |
| Suspension of payment | A measure to freeze a fraudster's account so that money cannot be withdrawn. Apply immediately through the financial institution's call center. (Chapter 5) |
| Application for remedy for damage | Applying in writing for damage relief at a financial institution's branch office after a suspension of payment. If payment suspension was requested by phone, the application must be submitted within three business days from that date (Article 3, Paragraph 2 of the Enforcement Decree of the Special Act on the Prevention of Loss Caused by Telecommunications-based Financial Fraud and Refund for Loss). (Chapter 5) |
| Indication of AI-generated content | A label or indication notifying that an output was created with generative AI. Article 31 of the AI Basic Act obligates business operators to provide this. (Chapter 5) |
| AI business operator | A legal entity, organization, individual, state agency, etc., engaged in business related to the artificial intelligence industry who corresponds to either a developer-operator that develops and provides artificial intelligence (Sub-item (a)) or a user-operator that provides products or services using that artificial intelligence (Sub-item (b)) (Article 2, Item 7 of the AI Basic Act). (Chapter 5) |
| User | A person who is provided with artificial intelligence products or services (Article 2, Item 8 of the AI Basic Act). (Chapter 5) |
| Fake review | A fabricated review made without actual user experience. Similar sentences and a sudden rush of praise are warning signs. (Chapter 5) |