🌅 Tuesday Morning — AI & Machine Learning English

Date: March 31, 2026 | Session: Morning | Level: Intermediate Start your day strong — 20 minutes of focused English practice on the language of AI.

---

🔤 Word of the Day: Hallucinate

IPA Pronunciation	/həˈluː.sɪ.neɪt/
Part of Speech	verb
Vietnamese Meaning	ảo giác; (trong AI) bịa đặt thông tin không có thật

How to Say It

Break it into 4 syllables: ha – LOO – si – nate Stress falls on the second syllable: ha-LOO-si-nate The “a” in the first syllable is a short schwa: /hə/ (like a quick “huh”)

3 Example Sentences

1. General meaning: “After three days without sleep, she started to hallucinate and saw things that weren’t there.”

2. AI/ML context: “The chatbot hallucinated a fake research paper and even invented a convincing-sounding author name.”

3. Professional context: “We need to add guardrails to prevent the model from hallucinating incorrect financial figures in client reports.”

🔗 Learn More

• 📖 Cambridge Dictionary — hallucinate
• 🎙 YouGlish — hear “hallucinate” in real speech
• 🤖 Andrej Karpathy explains AI hallucination (YouTube)

---

📋 Vocabulary Table: AI & LLM Essentials

Phrase	Vietnamese	Example Sentence
large language model (LLM)	mô hình ngôn ngữ lớn	“GPT-4 and Claude are both large language models trained on billions of tokens.”
prompt engineering	kỹ thuật viết prompt	“Good prompt engineering can dramatically improve the quality of AI-generated output.”
inference	suy luận / thực thi mô hình	“Running inference on this model requires at least 24 GB of VRAM.”
retrieval-augmented generation (RAG)	tạo sinh tăng cường truy xuất	“We use RAG to ground the chatbot’s answers in our internal documentation.”
fine-tuning	tinh chỉnh mô hình	“We fine-tuned the base model on our customer support data to improve accuracy.”

---

🗣 Pronunciation Guide

Deep Dive: hallucinate /həˈluː.sɪ.neɪt/

Syllable	Sound	Tip
ha-	/hə/	Relax your mouth — it’s a weak “huh” sound, not “hay”
-LOO-	/luː/	Stressed! Round your lips like you’re saying “loom”
-si-	/sɪ/	Quick and light, like “sit” without the “t”
-nate	/neɪt/	Clear and strong — rhymes with “gate” or “plate”

Common Mistakes for Vietnamese speakers

• ❌ ha-lu-si-NAY-t (wrong stress)
• ✅ ha-LOO-si-nate (stress on second syllable)
• Vietnamese speakers often add stress to the last syllable — resist this habit!

🎯 Practice Sentence (read aloud 3 times)

“The language model hallucinated a confident answer, citing a paper that had never been published.”

Focus on:

1. Stress on hall-LOO-cin-nated (secondary stress on -nat-)
2. Natural linking: “lan-gwidge mo-dell” (link words smoothly)
3. Falling intonation at the end — it’s a statement of fact, not a question

---

✏️ Exercise 1: Fill in the Blank

Choose the correct word: hallucinate / inference / fine-tune / prompt / RAG

1. We need to ________ the model on Vietnamese customer queries to improve its accuracy.
2. The engineer wrote a detailed ________ to get the AI to summarize meeting notes in bullet points.
3. During testing, the model began to ________ — it confidently cited statistics that don’t exist.
4. We added a ________ pipeline so the chatbot retrieves answers from our internal wiki before generating a response.
5. Running ________ on a 70-billion-parameter model is too slow without GPU acceleration.

✅ Click to reveal answers

1. fine-tune — you adjust the model’s weights for a specific domain
2. prompt — the input instruction given to an LLM
3. hallucinate — the model generates false but plausible-sounding content
4. RAG (retrieval-augmented generation) — grounds the model in real documents
5. inference — the process of running a trained model to get predictions

---

✏️ Exercise 2: Translate to English

Translate these Vietnamese sentences into natural English. Use today’s vocabulary.

1. “Mô hình đã bịa ra một tài liệu tham khảo giả — đây là một ví dụ điển hình của hallucination trong AI.”
2. “Chúng tôi đang tinh chỉnh mô hình cơ sở trên dữ liệu nội bộ để cải thiện hiệu suất.”
3. “RAG cho phép mô hình truy xuất thông tin thực tế trước khi tạo ra câu trả lời.”

✅ Click to reveal suggested answers

1. “The model fabricated a fake reference — this is a classic example of AI hallucination.” (or: “…this is a textbook case of hallucination in AI.”)

2. “We are fine-tuning the base model on our internal data to improve performance.” (or: “We’re fine-tuning the foundation model on proprietary data to boost accuracy.”)

3. “RAG allows the model to retrieve factual information before generating a response.” (or: “With RAG, the model fetches real context first, then generates a grounded answer.”)

---

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

Vietnamese meaning	“đầu vào rác, đầu ra rác” — chất lượng đầu ra phụ thuộc vào chất lượng đầu vào
Used in	AI/ML, data engineering, software development
Tone	Professional, matter-of-fact

Example 1:

“We spent weeks debugging why the model’s predictions were so poor — turns out, garbage in, garbage out. The training data was full of duplicates and mislabeled examples.”

Example 2:

“Your prompt engineering matters, but remember: garbage in, garbage out — if you give the model vague instructions, you’ll get vague answers back.”

---

📺 Recommended Watching

Level up your AI English and your technical knowledge at the same time:

Channel / Video	Why Watch It	Level
Andrej Karpathy — Let’s build GPT	Deep technical English, clear explanations of LLM internals	Advanced
AI Explained	Fast-paced, current AI news — great for listening comprehension	Intermediate
Yannic Kilcher	Paper walkthroughs with natural academic English	Advanced

💡 Tip: Watch at 0.75x speed first, then 1x. Pause and repeat any sentence you didn’t catch fully.

---

🎯 Today’s Daily Challenge

Action: Open your current project (or any AI tool you use) and write a prompt for a task you normally do manually. Then say the prompt out loud before you type it.

Ask yourself:

• Am I using the right technical vocabulary?
• Did I give the model enough context (who, what, format, tone)?
• Would a non-native English speaker understand this prompt?

“The quality of your prompt is a direct reflection of the clarity of your thinking.”

---

🔁 Quick Review

What you learned today	✅
Word: hallucinate — AI generates false information	☐
Phrase: large language model, prompt engineering, RAG, inference, fine-tuning	☐
Idiom: garbage in, garbage out	☐
Pronunciation: stress on ha-LOO-si-nate	☐
Practice sentence read aloud 3×	☐

---

🌅 Great start to Tuesday! Come back this evening for the evening session — we’ll dive deeper into AI conversation scenarios and listening practice.

Part of the Daily English series for tech professionals.