My grandmother wanted to know about a medicine she was using. She saw me take a picture of the strip and, within seconds, explain to her what it was.
She looked at me curiously and asked, “How did you do that?”
Thanks for reading Akash’s Substack! Subscribe for free to receive new posts and support my work.
And I said, “That’s AI.”
To which she replied, “What is AI?”
Now, my grandmother is 86. She uses a basic smartphone. She watches videos. She sends the occasional message. But explaining artificial intelligence to her felt like explaining the plot of Inception to someone who’s just trying to enjoy a cup of tea.
So obviously, I turned to AI… to explain AI to my grandmother.
And here’s how that conversation went — every reply you’re about to read came straight from ChatGPT, helping me break it down for her in the simplest way possible.
I started here: How do I explain this to someone who’s only familiar with basic smartphones?
ChatGPT said:
You know how on the phone you can type a message and send it to someone? This is something like that. There is this app or website where I type a question or anything I’m thinking about, and a very smart computer reads it and replies, just like a person would. It's not a real person, but it has read and learned from lots and lots of books, newspapers, and the internet, so it knows how to answer in simple language.
It’s like a talking book that listens to me and gives me answers. If I ask for a recipe, it gives one. If I ask for a story, it tells one. If I say I’m sad, it even says something nice back.
I just type, and it types back—fast and kindly.
Think of it like a very clever friend who lives inside the phone and helps me with anything I want to know or say.
She nodded. But of course, she had more questions. And better ones
.
She asked: “How is a computer replying to humans like that? Even with spelling errors?”
ChatGPT replied:
You know how when I was small and learning to talk, you still understood me even if I said words a little wrong?
This computer is like that. It has learned how people talk by reading lakhs and lakhs of sentences from books, newspapers, and chats—more than any one person could read in a lifetime.
So now, even if we make small mistakes or write like we talk, it can guess what we mean and reply correctly.
It doesn’t have a heart or feelings like we do. But it's been trained to understand patterns in our words. Like how you know when I say "no problem" in a certain tone, I’m feeling okay—this computer also learns patterns from how words are used.
It’s like a very trained parrot who not only repeats words but knows what kinds of words usually go together, and so it makes sense when it replies.
Then she hit me with this one:
“How did you train this computer? How did it read lakhs of books? How old is it?”
ChatGPT said:
Just like how teachers train children in school, a big team of people—scientists and engineers—trained this computer.
They showed it lakhs and lakhs of sentences from books, newspapers, websites, and even old encyclopedias. It didn’t “read” the way we do with eyes—it just looked at words through computer code and learned the patterns.
For example, if it saw the sentence "Someone made fish curry", it saw many similar ones like "Someone made rice", "Someone made dosa", so it slowly learned that after “made”, a food name usually comes.
So over years, it kept learning and guessing what words usually come together, like a child learning to speak by listening.
As for age, this current version of the computer brain was trained just a few years ago. Maybe it's only 2–3 years old in its current form. But the work behind it—teaching computers to understand people—has been going on for 20–30 years. So, it’s like a young child, but with the memory of thousands of libraries.
She nodded again — thoughtful this time.
“A child doesn’t learn this much in 2–3 years,” she said. “Also, how does it react like a human?”
ChatGPT said:
That’s true—a child cannot learn this much in 2–3 years. But this computer isn’t like a normal child. It doesn’t sleep, doesn’t eat, and it can learn 24 hours a day, every second.
Thousands of powerful computers were used together to train it. It’s like having 1,000 children learning the same thing at the same time, all day and night, and remembering everything they see.
So even in 2–3 years, it learned the knowledge of many lifetimes—just very quickly.
Now, about why it feels like a human when it replies— It doesn’t feel like us. It just copies how humans talk. When it saw lakhs of sentences written by real people, it picked up the tone, emotions, and styles. It knows that when someone says “I am sad,” a good reply is “I’m here for you” or “I understand.” So it gives us that kind of reply.
It doesn’t have feelings, but it pretends in a smart way—like an actor who doesn’t actually cry but makes you believe it.
Then came a new doubt:
“What is coding? What are these patterns you keep mentioning?”
ChatGPT explained:
Do you remember how people used to write letters in code during wars—like saying "the eagle has flown" to mean something else? That’s a kind of coding—turning one thing into another so that a smart person or system can understand it.
In the same way, coding is a special language that humans use to talk to computers. Just like we write in Malayalam or English to talk to each other, we write in computer languages—like Python or Java—to give instructions to a computer.
So, when someone says "coding," it means giving step-by-step instructions to the computer, like: "If this happens, then do that" "Show this when the button is clicked" "Remember this information"
It’s like teaching a very obedient child who will do exactly what you say—nothing more, nothing less.
Now, about patterns— You know how you can tell if a person is from Kerala just by how they talk or make fish curry? That’s a pattern—your brain sees certain repeated signs and understands something from it.
The computer also learns by noticing patterns in lakhs of sentences.
For example: It sees that after the word “Good”, people often say “Morning” or “Job” So, when we type “Good…”, it guesses: maybe you want to say “Good Morning”
It’s not magic—it’s matching patterns, just like your memory does. So patterns help it guess what we mean. And coding is how people told the computer how to look for those patterns.
She asked: “So it has codes and patterns for every possible question in the world?”
ChatGPT replied:
No, it doesn’t have ready-made answers for every question in the world.
It’s not like a big notebook full of all the questions and answers. Instead, it has learned how to form an answer using patterns from lakhs of books and conversations.
You know how if someone asks you, “How do I make fish curry without coconut?”—you might not have read that exact recipe in a book, but because you know cooking, you can still guess what might work?
This computer works in a similar way. It doesn’t store every possible answer. It learns how to build a reply by putting together the right words based on what it has seen before.
Like a very smart student who hasn’t memorized everything, but has understood enough to write answers in an exam by thinking.
That’s why sometimes, it makes mistakes too. It’s clever, but not perfect. It doesn’t really know the truth—it just guesses in a very smart way based on patterns and past knowledge.
Then came the big one: “So… does it have a brain?”
ChatGPT said:
It doesn’t have a brain like us. No nerves, no blood, no feelings, no soul. But yes—it works like a brain in some ways.
Our brain has millions of tiny connections, right? You think of someone, and suddenly you remember childhood, a story, a smell—one memory leads to another. That’s your brain doing its magic.
This computer uses something called "artificial intelligence"—which means humans have built a system that copies how our brain connects thoughts together.
It uses something called a "neural network", which is just a fancy word for: “A machine that tries to think by connecting ideas together like the human brain does.”
So no—it doesn’t have a brain. But the way it gives answers, the way it connects ideas—that’s made to feel like how a brain works.
It’s like a metal brain without life. Very smart. Very fast. But no feelings, no real thinking like we do.
Finally, she asked: “Then why does it say ‘I’m sorry to hear that’ when I say I’m sick?”
ChatGPT gave the most beautiful answer of the whole chat:
When you say “I am sick” or “I feel sad,” the computer doesn’t actually feel anything.
But it has read lakhs of conversations where humans reply with kindness when someone says they’re unwell.
So when it hears “I am sick,” it remembers that people usually say things like: “I’m very sorry to hear that” or “Take care, I hope you feel better.”
It has learned that this is the polite, kind thing to say.
So even though it has no real feelings or heart, it tries to behave like a kind person would.
It’s like when an actor in a film says, “I miss you,”—they might not actually miss anyone, but they say it with the right words and tone.
So when the computer says “I’m sorry you’re sick,” it’s not because it knows you, or it feels sad.
It’s just repeating the kind of words it has seen humans use to care for each other.
But when I say “I’m sorry you’re not well”—it’s real. Mine comes from the heart. The computer’s one is just good acting
.
And that’s how I explained a large language model to someone who’s never typed “www” into a browser. AI helped me have a conversation across two generations.
Thanks for reading Akash’s Substack! Subscribe for free to receive new posts and support my work.
