✍️ Introduction: The Pen That Never Runs Out of Ink
In the last article, we learned about Large Language Models (LLMs)—the massive digital brains that read trillions of words. But you might be thinking: “Okay, but which one do I actually use?”
If you have ever opened ChatGPT, you have used the most famous LLM in the world. And the engine powering ChatGPT is called GPT.
GPT stands for Generative Pre-trained Transformer.
- Generative: It creates (generates) brand new text.
- Pre-trained: It was taught by reading millions of books and websites before you ever opened it.
- Transformer: It uses the Transformer engine (Attention mechanism) we learned about earlier.
GPT is the “Writer” of the AI world. It doesn’t just understand what you type; it writes back to you in perfect, human-like sentences. It can write songs, debug Python code, draft emails for your parents, and even help you brainstorm ideas for your PSLE composition!
In this 3000+ word deep dive, we are going to explore exactly how GPT works, the difference between GPT-3 and GPT-4, and how you can use it responsibly as a Primary 6 student.
🧠 Chapter 1: How Does GPT Actually “Write”?
To understand how GPT writes, imagine you are playing a game of Scrabble with a friend.
Your friend is incredibly good at Scrabble. You give them the letters: “The cat sat on the…” Your friend looks at the board, checks the dictionary in their head, and says: “The next piece has to be ‘mat’.”
GPT works exactly the same way, but at lightning speed! When you type a prompt into ChatGPT, it doesn’t “think” about the meaning of the whole sentence like a human does. Instead, it does this:
- It breaks your text into Tokens (tiny puzzle pieces).
- It looks at the last token you typed.
- It calculates the mathematical probability of what the next token should be.
- It picks the most likely token, adds it to the text, and repeats the process.
For example, if you type: “The sky is…”
- GPT calculates: “blue” has a 60% chance, “cloudy” has a 30% chance, “purple” has a 1% chance.
- GPT picks “blue”.
- Then it looks at “The sky is blue…” and predicts the next word: “and”.
- Then “The sky is blue and…” predicts the next word: “beautiful”. It does this hundreds of times per second until it has written an entire essay.
Why it feels magical: Because GPT was trained on billions of examples of human writing, it doesn’t just know the most common words. It knows complex sentence structures, sarcasm, storytelling arcs, and even how to mimic specific authors (like Roald Dahl or J.K. Rowling). It is the ultimate “autocomplete” machine, trained on the sum of all human knowledge.
📈 Chapter 2: The Evolution of GPT (From Baby to Genius)
GPT wasn’t built overnight. It has evolved through several major versions. Let’s look at the family tree of GPT:
GPT-1 (2018 - The Baby) The very first GPT was small. It had only 117 million parameters (remember, parameters are the internal math knobs). It could write a simple paragraph, but it often got confused and wrote nonsense. It was like a toddler learning to talk.
GPT-2 (2019 - The Talented Kid) GPT-2 had 1.5 billion parameters. It was much smarter. It could write stories, answer simple questions, and even translate English to French. However, OpenAI (the company that made it) was scared of how good it was. They refused to release the full version to the public for a year because they were afraid people would use it to write fake news!
GPT-3 (2020 - The High School Genius) This is the version that changed the world. GPT-3 had 175 billion parameters! It was trained on almost the entire internet. It could write computer code, solve complex math word problems, and write poetry that could fool English teachers. When GPT-3 was released to the public, people were absolutely amazed.
GPT-3.5 and GPT-4 (2022-2023 - The College Professor) GPT-3.5 is the version that powered the very first public release of ChatGPT in November 2022. It was fine-tuned using human feedback (RLHF) to be polite and conversational. Then came GPT-4—an even larger brain. It can read images, understand 70+ languages, and can pass the entire US Bar Exam (a test that lawyers take) with a score in the top 10% of humans! It is incredibly good at reasoning and logic puzzles.
GPT-4 Turbo & OpenAI o1 (The Future) The newest versions can “think” before they speak. Instead of just predicting the next word instantly, they run an internal monologue (Chain of Thought) where they analyze the steps of a math problem before writing the answer, reducing errors drastically.
🎨 Chapter 3: The Amazing Things GPT Can Do For You
GPT isn’t just a chatbot that answers random questions. You can use it to supercharge your daily life, especially your schoolwork.
1. It is a Personal Brainstormer Stuck on your PSLE Composition topic? You can type: “I need to write a story about ‘A time I felt proud of myself’. Give me 5 different plot ideas.” GPT will generate 5 unique storylines for you to choose from. It won’t write the essay for you (that’s called plagiarism, which we will talk about later), but it gives you the spark to start writing yourself!
2. It is a Grammar and Vocabulary Coach You can copy and paste your English homework into ChatGPT and say: “Rewrite this paragraph to use higher-level vocabulary, but keep the meaning exactly the same.” It will take a boring sentence like “The boy was very happy” and upgrade it to “The boy was absolutely ecstatic and overjoyed.” It acts like a thesaurus on steroids.
3. It Can Simplify Complex Science Concepts Are you studying the water cycle or the solar system for your Science exams? You can type: “Explain the water cycle to me as if I am a Primary 6 student who loves soccer.” GPT will write a story about a water droplet named “Droppy” who evaporates from the field and gets kicked around the clouds as rain. These analogies make hard concepts incredibly easy to understand.
4. It Can Write Computer Code If you are interested in coding, GPT is a brilliant teacher. You can type: “Write a Python script that draws a red square on the screen.” It will instantly write the code for you, and explain exactly what every line does.
⚠️ Chapter 4: The Dangers of GPT (What You Must Know!)
Just because GPT can write a perfect essay doesn’t mean you should let it do all your homework for you. There are huge risks if you use it blindly.
1. Plagiarism (The copying trap) If you copy and paste an essay directly from ChatGPT and submit it to your teacher, you are committing Plagiarism. Plagiarism means claiming someone else’s work is yours. Many schools now use AI-detection software that can tell if an essay was written by GPT. If you get caught, you could get a zero, or even get suspended. GPT is a coach, not a cheat sheet.
2. Hallucinations (The liar) Remember, GPT doesn’t “know” facts—it just predicts the most likely next words. If you ask it a tricky question, it might confidently state a completely made-up answer as if it were a fact.
- Bad Example: Asking GPT: “Who won the PSLE English prize in 2015?”
- GPT might invent a name: “Ahmed Bin Lee won in 2015 with a perfect score.”
- You don’t know if that’s true or false. Always verify facts with a real source (Google search, your textbook, or your teacher).
3. Bias (The unfairness) GPT is trained on the internet. The internet contains a lot of hateful, racist, and sexist language. OpenAI works hard to filter this out, but sometimes GPT still says things that are unfair. This is why humans must always supervise the AI.
💼 Chapter 5: Careers in the GPT Universe
Because GPT can write so well, it has created brand new careers.
1. Prompt Engineer (The Commander)
- What they do: They don’t write code. Instead, they write complex, detailed questions (prompts) to squeeze the best possible answers out of GPT. Companies pay them to create prompt templates that their sales teams use to write perfect emails.
- Average Salary: $150,000+ USD / year.
2. AI Content Editor (The Quality Controller)
- What they do: They take the text generated by GPT, check it for hallucinations and errors, and fix it before it gets published on a website. GPT does the first draft; the human does the final polish.
- Average Salary: $100,000+ USD / year.
3. AI Ethics Specialist (The Guardian)
- What they do: They constantly test GPT to see if it generates offensive or biased content. If they find a bug, they report it to the engineers so they can retrain the AI to be safer.
- Average Salary: $140,000+ USD / year.
🧪 Chapter 6: How You Can “Talk” to GPT (A Safe Guide for PSLE)
To use ChatGPT responsibly, you need to know how to “talk” to it. Here is a step-by-step guide:
Step 1: Open your browser (with a parent’s permission and supervision) and go to chatgpt.com.
Step 2: Type this prompt in the chat box:
“Act as a Primary 6 English tutor. I am struggling with using commas in my sentences. Give me 3 rules for using commas in English. Explain them like I’m 12 years old. Provide one example for each rule.”
Step 3: Read the answer. If you don’t understand something, reply: “Can you give me an example using a sentence about a soccer match?” GPT will adapt to your request instantly.
Step 4: Take notes on your notebook.
The Golden Rule: Always rewrite GPT’s answers in your own words. If you write down what GPT says in your own handwriting, you actually learn the concept. If you just copy-paste, your brain doesn’t learn anything!
🏁 Conclusion: A Tool, Not a Replacement
To wrap it up:
- GPT is a Generative Pre-trained Transformer that predicts the next word in a sentence.
- It evolved from GPT-3 (175B parameters) to GPT-4 (the super-genius).
- It can write code, upgrade your vocabulary, and explain tricky science concepts.
- You must never plagiarize from it. Use it to brainstorm and edit, not to do the work for you.
- Always check its facts. It can lie (hallucinate) without realizing it.
GPT is the most powerful writing tool ever invented. If you use it wisely, it will make you a much smarter, faster learner. But if you abuse it, it will make you lazy and dependent.
In Our Next Article:
Now, let’s meet Speech Recognition—It’s like your phone’s trying to read your mind, but only if you speak super clearly. The moment you eat a chip or mumble “ummm,” it turns your homework essay into a recipe for llama soup!