🌅 Tuesday Morning — AI & Machine Learning English

Good morning! Today we dive into the language of Artificial Intelligence and Machine Learning — one of the hottest domains in tech. Mastering this vocabulary will help you read papers, join discussions, and present your AI work confidently in English.

---

🔤 Word of the Day: Inference

Word	inference
IPA	/ˈɪn.fər.əns/
Part of Speech	noun
Vietnamese	suy luận / giai đoạn dự đoán (trong AI)

📖 What does it mean in AI?

In machine learning, inference is the process of using a trained model to make predictions on new data — as opposed to training, which is when the model learns from data.

Think of it this way: training = studying for an exam. Inference = actually taking the exam.

✏️ Example Sentences

1. “We run inference on the edge device to reduce latency.” → Chúng tôi chạy inference trên thiết bị biên để giảm độ trễ.

2. “The model’s inference speed improved after we switched to a quantized version.” → Tốc độ inference của mô hình được cải thiện sau khi chúng tôi chuyển sang phiên bản lượng tử hóa.

3. “During inference, the LLM generates tokens one at a time in an autoregressive manner.” → Trong quá trình inference, LLM tạo token từng cái một theo cách hồi quy.

🔗 Resources

• 📖 Cambridge Dictionary — inference
• 🎧 YouGlish — hear “inference” used in real speech
• 🎬 Andrej Karpathy explains LLM inference

---

📚 Vocabulary Table: AI & Machine Learning Phrases

Phrase	Vietnamese	Example Sentence
hallucination	ảo giác (AI bịa đặt thông tin)	“The model produced a hallucination — it cited a paper that doesn’t exist.”
fine-tuning	tinh chỉnh mô hình	“We fine-tuned GPT-4o on our internal dataset to improve domain accuracy.”
retrieval-augmented generation (RAG)	sinh văn bản có tăng cường truy xuất	“RAG helps reduce hallucinations by grounding the model in real documents.”
context window	cửa sổ ngữ cảnh	“This model has a 128k token context window, so it can handle long documents.”
prompt engineering	kỹ thuật thiết kế câu lệnh	“Good prompt engineering can dramatically improve output quality without retraining.”

---

🗣️ Pronunciation Guide

Breaking down: “inference”

IN  -  fer  -  ence
/ɪn/  /fər/  /əns/

• IN → short “i” as in it, is, in — NOT “eye-n”
• fer → weak syllable, like “fur” with a schwa — /fər/
• ence → /əns/ — soft ending, almost swallowed

🎯 Common mistake: Vietnamese speakers often say “in-FER-ence” (stressing the second syllable). The stress falls on the first syllable: IN-fer-ence.

🔊 Practice Sentence — Read aloud 3 times:

“Fast inference on large language models requires optimized hardware and quantization techniques.”

Round	Focus
1st read	Slow — focus on each word’s pronunciation
2nd read	Normal speed — natural rhythm
3rd read	Confident — as if presenting to your team

Key words to nail:

• inference → /ˈɪn.fər.əns/
• language → /ˈlæŋ.ɡwɪdʒ/ (not “lan-guage”)
• quantization → /ˌkwɒn.tɪˈzeɪ.ʃən/
• techniques → /tekˈniːks/

---

✍️ Exercise 1: Fill in the Blank

Choose the correct word: inference / hallucination / fine-tuning / RAG / context window

1. “The ________ of this model is only 4,000 tokens, so it can’t process long PDFs.”
2. “To prevent the chatbot from making up facts, we implemented a ________ pipeline with a vector database.”
3. “The sales demo failed because the model produced a ________ — it invented a product feature that doesn’t exist.”
4. “We did ________ on a dataset of customer support tickets to make the model more helpful.”
5. “The ________ endpoint is exposed via REST API and handles about 500 requests per second.”

✅ Click to see answers

1. context window
2. RAG (retrieval-augmented generation)
3. hallucination
4. fine-tuning
5. inference

---

🔄 Exercise 2: Translate into English

Translate these Vietnamese sentences into natural English. Think about how a senior AI engineer would say this in a team meeting.

1. “Mô hình này bị ảo giác rất nhiều khi được hỏi về dữ liệu thời gian thực.”
2. “Chúng ta cần tinh chỉnh mô hình trên tập dữ liệu nội bộ của công ty.”
3. “Cửa sổ ngữ cảnh 128k token cho phép chúng ta truyền toàn bộ codebase vào prompt.”

✅ Click to see suggested answers

1. “This model hallucinates a lot when asked about real-time data.” → Or: “The model tends to hallucinate frequently on real-time queries.”

2. “We need to fine-tune the model on the company’s internal dataset.”

3. “The 128k token context window lets us pass the entire codebase into the prompt.”

---

💡 Idiom of the Day: “garbage in, garbage out”

Idiom	garbage in, garbage out
Abbreviation	GIGO
Vietnamese	“rác vào, rác ra” — nếu dữ liệu đầu vào tệ, kết quả đầu ra cũng tệ

This classic computing phrase is heavily used in AI/ML discussions to explain why data quality matters as much as model architecture.

📌 Examples in context:

1. “Our model’s predictions were terrible last quarter — classic garbage in, garbage out. We hadn’t cleaned the training data properly.” → Dự đoán của mô hình rất tệ trong quý vừa rồi — điển hình của garbage in, garbage out. Chúng tôi đã không làm sạch dữ liệu huấn luyện đúng cách.

2. “Before we blame the LLM, let’s audit our prompts — garbage in, garbage out applies to prompt engineering too.” → Trước khi đổ lỗi cho LLM, hãy kiểm tra lại prompt của chúng ta — garbage in, garbage out cũng áp dụng cho prompt engineering.

---

📺 Recommended Watching

Level up your AI English with these channels:

Resource	Why Watch	Link
Andrej Karpathy	Former Tesla/OpenAI engineer; clear, deep explanations of LLMs	youtube.com/@AndrejKarpathy
Yannic Kilcher	ML paper walkthroughs with technical English commentary	youtube.com/@YannicKilcher
AI Explained	Accessible AI news and model breakdowns — great for listening practice	youtube.com/@aiexplained-official

🎯 Tip: Watch with English subtitles (not Vietnamese). Pause and repeat sentences you find hard to follow.

---

🌄 Today’s Challenge

“One tiny action, done consistently, beats big plans done rarely.”

Your challenge for today:

In your next Slack message or PR description, use the word “inference” correctly in context.

For example:

• “This endpoint handles model inference — we should add a timeout.”
• “Inference latency increased after the model update. Investigating now.”

If you don’t have a chance at work, write one sentence in a comment on any AI-related GitHub issue or tweet/post.

⏱ Takes less than 2 minutes. Do it before lunch.

---

📊 Quick Recap

Item	What You Learned
🔤 Word	inference /ˈɪn.fər.əns/ — suy luận / giai đoạn dự đoán
📚 Phrases	hallucination, fine-tuning, RAG, context window, prompt engineering
💡 Idiom	garbage in, garbage out — rác vào, rác ra
🗣️ Pronunciation	Stress on IN-fer-ence, not in-FER-ence
✍️ Practice	2 exercises covering fill-in-blank and translation

---

See you this evening for the Afternoon session! 🚀

— Your English Coach at luonghongthuan.com