Tuesday Morning — AI & Machine Learning English
Good morning! Today we dive into one of the hottest areas in tech — AI and Machine Learning vocabulary. Whether you are explaining a chatbot to your manager or reading papers about LLMs, these terms will help you communicate confidently.
Word of the Day: prompt
IPA Pronunciation: /prɒmpt/
Vietnamese meaning:
- câu lệnh / câu hỏi cho AI — the text input you give to an AI model
- gợi ý / nhắc nhở — a cue or signal that triggers a response
Quick tip: Say it as one syllable: “PROMPT”. The “o” sounds like the “o” in “pot” or “hot”. End clearly with the “pt” cluster — do not drop the final “t”.
Cambridge Dictionary: prompt — Cambridge
Learn more: Prompt Engineering Tutorial — LearnPrompting.org
Example Sentences
- “She wrote a detailed prompt asking the model to summarise the meeting notes in three bullet points.”
- “If your prompt is too vague, the AI will give you a generic answer — be specific about the format and tone you want.”
- “Our team created a library of reusable prompts so every developer follows the same structure when calling the LLM.”
Vocabulary Table: AI & Machine Learning Terms
| Phrase | Vietnamese | Example |
|---|---|---|
| RAG (Retrieval Augmented Generation) | Sinh văn bản có truy xuất tài liệu | ”We built a RAG pipeline so the chatbot answers questions using our internal documentation rather than guessing.” |
| context window | cửa sổ ngữ cảnh — lượng văn bản model có thể đọc cùng lúc | ”The model has a 128k-token context window, meaning it can process a very long document in a single call.” |
| system prompt | câu lệnh hệ thống — hướng dẫn ẩn gửi trước khi user nhắn tin | ”The system prompt tells the assistant to always reply in formal English and never reveal its internal instructions.” |
| zero-shot | không có ví dụ mẫu — model trả lời ngay mà không cần ví dụ | ”We tested the model zero-shot first; it performed surprisingly well without any examples.” |
| fine-tuning | tinh chỉnh mô hình — huấn luyện thêm trên dữ liệu riêng | ”After fine-tuning the model on our customer support tickets, it handles domain-specific questions much better.” |
Pronunciation Guide: Tricky AI Terms
Let’s break down “prompt” and two commonly mispronounced AI words:
prompt /prɒmpt/
| Sound | Symbol | Tip |
|---|---|---|
| pr | /pr/ | Lips together for “p”, roll straight into “r” — no pause |
| o | /ɒ/ | Short, open “o” — like “pot”, “hot”, “lot” |
| mpt | /mpt/ | Three consonants together — keep them all: “m” then “p” then “t” |
Common mistake: Vietnamese speakers often say /pɾom/ and drop the final “pt”. Practise ending strongly: “promp-T”.
retrieval /rɪˈtriːvəl/
| Part | Sound | Like |
|---|---|---|
| re | /rɪ/ | short “ri” — like “rip” |
| trie | /triːv/ | long “ee” — “tree” + “v” |
| al | /əl/ | unstressed schwa + “l” |
Stress: re-TRIEV-al. The middle syllable carries the weight.
augmented /ɔːɡˈmentɪd/
| Part | Sound | Like |
|---|---|---|
| aug | /ɔːɡ/ | like “org” in “organ” — long “aw” sound |
| men | /ˈment/ | stressed — like “ment” in “mental” |
| ed | /ɪd/ | light “id” ending |
Stress: aug-MENT-ed.
Practice Sentence — Read 3 Times Aloud
“I fine-tuned my prompt to improve the model’s retrieval-augmented responses.”
Read slowly the first time, then at a normal pace. Focus on:
- prompt — short “o”, clear final “t”
- re-TRIEV-al — stress the middle
- aug-MENT-ed — stress the second syllable
Exercise 1: Fill in the Blank
Choose the correct word: prompt / RAG / context window / system prompt / zero-shot
- The engineer added a ________ that instructed the assistant to only answer questions related to software development.
- Without any examples, the model solved the problem ________, which surprised the whole team.
- Because the document was too long for the model’s ________, we had to split it into smaller chunks.
- A good ________ is specific, includes context, and states the desired output format clearly.
- The company built a ________ system so the AI can search internal wikis before generating an answer.
Show Answers
- system prompt
- zero-shot
- context window
- prompt
- RAG
Exercise 2: Translate to English
Translate these Vietnamese sentences about AI prompting into natural English.
- “Tôi đã viết một câu lệnh chi tiết để yêu cầu AI tóm tắt cuộc họp trong năm điểm chính.”
- “Mô hình được tinh chỉnh trên dữ liệu của chúng tôi nên nó trả lời chính xác hơn nhiều.”
- “Hệ thống RAG giúp chatbot truy xuất tài liệu nội bộ trước khi trả lời người dùng.”
Show Answers
- “I wrote a detailed prompt asking the AI to summarise the meeting in five key points.”
- “The model was fine-tuned on our data, so it answers much more accurately.”
- “The RAG system helps the chatbot retrieve internal documents before responding to the user.”
Idiom of the Day: “think outside the box”
Meaning: to think creatively and come up with ideas that go beyond the usual approach
Vietnamese: suy nghĩ sáng tạo, vượt ra ngoài khuôn mẫu thông thường
This idiom fits perfectly in prompt engineering — the best prompts often come from creative framing, not just following templates.
AI-context examples:
- “To get the model to write better code comments, you have to think outside the box — try asking it to explain the code as if teaching a junior developer.”
- “Our team thought outside the box with the system prompt: instead of just listing rules, we gave the AI a persona, and the tone improved dramatically.”
Recommended Watching & Reading
- Andrej Karpathy — “Intro to LLMs” on YouTube: Clear, visual explanations of how language models work — ideal for developers. Search: “Andrej Karpathy Intro to LLMs”
- DAIR.AI — Prompt Engineering Guide: promptingguide.ai — comprehensive, free resource covering zero-shot, few-shot, chain-of-thought, and RAG prompting techniques.
- LearnPrompting.org: learnprompting.org — beginner-friendly, interactive lessons on writing effective prompts with real examples.
Daily Challenge: The Same Task, Three Different Prompts
Task: Ask an AI to summarise a meeting.
Write three different prompts for the same task and compare the outputs. Notice how the wording, tone, and structure change the result:
Prompt 1 — Basic:
“Summarise this meeting.”
Prompt 2 — With format instruction:
“Summarise the following meeting transcript in three bullet points. Focus on decisions made and action items assigned.”
Prompt 3 — With role and audience:
“You are a project manager. Summarise this meeting for a non-technical stakeholder in plain language. Highlight the most important decision and the next step.”
Try all three with any AI tool (ChatGPT, Claude, Gemini). Write down what changes between responses. This exercise builds your intuition for prompt engineering faster than any tutorial.
See you at noon for the next session — today’s midday lesson covers the word “hallucinate” in the context of AI errors. Evening session: “inference” in model deployment.