Talk to AI Like a Pro — The Filmy Guide to Prompt Engineering

Imagine AI is like your Bollywood hero—sometimes it understands your dialogues perfectly, and sometimes it gets all confused like a sidekick in a comedy scene. Prompt Engineering is all about how you talk to that hero (AI) so you get the blockbuster answers you want.

Let’s dive into this masaledaar world of prompt engineering with examples that will make you say, “Wah re AI, kya baat hai!”

Now it's time to pull back the curtain and meet the real villain of the story — no, not Mogambo, but something far more dangerous in the AI world: GIGO — Garbage In, Garbage Out.

Imagine this: you're trying to get ChatGPT to behave like Shah Rukh Khan, but you feed it a prompt like, “Hey, tell me something cool.” Bro... what is this? This is not a prompt; this is how a confused fresher asks for life advice at 3 AM.

Just like in Bollywood where bad scripts ruin even superstar films (looking at you, Race 3), in AI, bad inputs lead to useless outputs — no matter how powerful your model is.

So, let's dive into this villain’s lair and learn how to avoid giving our chatbot the digital version of biryani without salt.

---

1. GIGO — Garbage In, Garbage Out

Let’s get real. You wouldn’t expect a Bollywood movie to make sense if the dialogues were written during a hangover, right? Similarly, you can't expect your AI model to give epic answers when your inputs are a hot mess.

GIGO stands for Garbage In, Garbage Out — and it’s the golden rule of any AI system. If your prompt is vague, messy, or unclear, your AI output will be more lost than Kareena Kapoor in Jab We Met before Manali.

Example from Bollywood:

‣ Bad Prompt:
"Tell me about something."
This is like asking Salman Khan in Wanted, “Bhai, thoda dance dikha do.” You’re not giving any direction — result? Confused bhai energy.

‣ Good Prompt:
"Explain the concept of overfitting in machine learning with a Shah Rukh Khan movie plot."
Now we’re talking! This is precise, fun, and full of personality. Like a perfectly written Rohit Shetty climax.

• Moral of the story?
‣ ChatGPT, Claude, Gemini, LLaMA — sab bhale hi AI ke Baazigar ho... but if your prompt is trash, the output will be even trashier. So acha socho, acha likho, and your AI will return answers worthy of a standing ovation.

---

2. Input Tokens — The Dialogues of AI

Ever tried talking to ChatGPT like you're pouring your heart out in a Karan Johar movie?
And suddenly—"You've reached the token limit :("

That, my friend, is the reality of input tokens.

• What Are Input Tokens?

Think of tokens as currency units for LLMs. Every word, punctuation mark, emoji — even your “😂” when you laugh at your own joke — counts as a token. The model reads your input in tokens, not full words.

‣ Hello world! → 3 tokens

‣ Mahiiii maar raha hai → 5 tokens

‣ Example:

Imagine you're pitching a script to Yash Raj Studios:

‣ If you're crisp like “DDLJ in Europe + Family Drama + Train Scene” → Approved!

‣ If you're like “So there’s this guy who kinda loves this girl, and then there’s a cow, and also maybe a spaceship…” → Rejected. Too long, too messy.

LLMs work the same way. Keep it tight and punchy.

• Moral?
‣ Always write your prompts like you're scripting for Zoya Akhtar — minimal, poetic, and powerful. Tokens matter. Be token-wise, not token-wasteful!

---

3. Prompting: The Catalyst for Machine Brilliance.

‣ Imagine you're Shah Rukh Khan in Chennai Express trying to ask for directions in Tamil — and the only Tamil you know is “lungi dance”. That’s what a bad prompt feels like to an LLM — confused, lost, and slightly offended 😅

• What Is Prompting?

‣ Prompting is how you talk to a language model so it understands exactly what you want. It’s like being the director on set. Give unclear instructions, and your hero might dance during an emotional breakup scene.

Prompting = Directing the Model’s Behaviour

Examples:

• Bad Prompt: “Tell me something.”

• Good Prompt: “Write a 100-word motivational speech in the style of Shahrukh Khan from Chak De India.”

Now we’re talking!

---

4. Prompt Dialects: How AI Understands Alpaca, ChatML, and INST

‣ Just like Bollywood stars speak Hindi, English, Punjabi, and even “text-message-speak,” language models understand different prompt formats or "languages" that help them get you better.

Some popular formats in the prompting universe are listed below:

• Alpaca Prompt – The Friendly Neighbor

‣ Imagine Alpaca as that friendly guy who always helps you but in a neat Q&A style.

### Instruction:
Explain how a blockchain works.  

### Input:  
None  

### Response:  
Blockchain is like a ledger everyone shares...

This format is clear and structured, perfect for instruction-following models.

• ChatML Prompt – The Scriptwriter's Format

‣ OpenAI’s ChatML is like a screenplay for chat models — it gives the AI the entire dialogue history with roles clearly marked: system, user, assistant.

<|system|> You are a helpful assistant. <|end|>  
<|user|> Explain recursion. <|end|>  
<|assistant|> Recursion is when... <|end|>

It’s like having a perfect Bollywood script where every character knows their lines and when to speak!

• Inst Format (Instruction Format) – The Directorial Command

‣ Inst Format is like the strict director yelling, “Cut! Do it again, but this time make it funny!” It’s designed for direct instructions with input and expected output.

Instruction: Summarize the movie ‘3 Idiots’.  
Input: The story follows three friends...  
Output: ‘3 Idiots’ is about friendship, dreams...

Why Use Different Prompt Formats?

‣ Think of them as different dialects to talk to your AI. Some formats are better for certain models — like how Govinda’s comedy works best in masala movies, not in serious dramas!

Now that we’ve explored how prompts speak different “languages” — like Alpaca, ChatML, and Instruction formats — it’s time to dive into the how of prompting. Just like Bollywood heroes have their unique styles — action, comedy, romance — AI has different prompting techniques to get the best performance. Let’s explore these blockbuster methods that make your chatbot a star!

---

5. Prompting Techniques – The Absolute Cinema!

• Zero-Shot Prompting: The Fresh Hero Walk-in

‣ Imagine Shah Rukh Khan walking onto the set for the very first time with no rehearsals, no script—just pure instinct. That’s zero-shot prompting: you give the model a question or task without any examples, and it tries its best to answer. Sometimes it nails it, sometimes it fumbles, but that raw energy is exciting!

‣ Example:
“Translate this sentence into French.”
The model just jumps in and tries, no prior cues.

• Few-Shot Prompting: The Tutor’s Coaching Montage

‣ Like Amitabh Bachchan giving a pep talk to a newbie in Paa or Pink, few-shot prompting shows the model a few examples before asking it to perform. It’s like giving a quick tutorial so the AI knows the vibe and style.

‣ Example:
“Here are some English to French translations: ‘Hello’ → ‘Bonjour’, ‘Thank you’ → ‘Merci’. Now translate ‘Good morning’.”

• Chain-of-Thought Prompting: The Detective’s Thought Process

‣ Think of Sherlock Holmes or Raj in Andaz Apna Apna, talking aloud to connect clues. Chain-of-thought prompting makes the AI explain its reasoning step-by-step before answering. This technique boosts accuracy by revealing its internal logic.

‣ Example:
“To solve this math problem, first add these numbers, then multiply the result by...” and so on.

• Self-Consistency Prompting: The Bollywood Double Take

‣ Imagine a scene where the hero imagines multiple endings, then chooses the best one—like a Bollywood director shooting several versions of the climax. Self-consistency prompting has the AI generate multiple answers and pick the most consistent or frequent one.

• Persona-Based Prompting: Acting Like Your Favourite Star

‣ This is where your AI dons the avatar of Shah Rukh Khan or any persona you want! Like Ranveer Singh becoming Alauddin Khilji in Padmaavat, the chatbot mimics a personality’s tone, style, and knowledge. This makes conversations more engaging and personal.

---

6. Final Words

‣ Prompt Engineering is like directing your AI hero in a blockbuster movie. Give clear dialogues, set the mood, and use the right techniques to get award-winning performances.

So next time you chat with AI, remember: “Tumhara prompt hi tumhara superstar hai!”