7 shocking secrets about ChatGPT that will blow your mind

Artificial Intelligence (AI) has taken the world by storm, but few AI tools have sparked as much curiosity and conversation as ChatGPT.

Since its launch in 2022, it has become more than just a chatbot; it’s been a virtual writing assistant, problem-solver, teacher, and even an interview coach for millions around the globe.

Here are seven surprising facts about ChatGPT that might just change the way you look at it forever.

1. One of the fastest-growing apps in history

When ChatGPT launched, it didn’t just gain traction; it exploded. Within a few days, it attracted over one million users, and by November 2023, it was handling more than 10 million queries daily with 100 million weekly active users.

That pace of growth rivals, and even surpasses, some of the biggest tech launches in history.

2. Humans teach ChatGPT how to behave

Before it becomes the helpful assistant you know, ChatGPT goes through a process called pre-training, where it learns from massive amounts of text by predicting the next word in a sentence.

That’s where human feedback comes in. Trained “annotators” test the model, rank its responses, and align it with ethical guidelines.

3. ChatGPT thinks in tokens, not words

While humans learn through words and sentences, ChatGPT understands the world in tokens, smaller chunks of text that could be entire words, partial words, or even random character strings.

Sometimes, tokenisation can produce odd splits, which is why AI occasionally misunderstands certain phrases.

4. ChatGPT’s knowledge gets outdated everyday

ChatGPT’s intelligence isn’t updated in real-time. Instead, it has a knowledge cutoff date, meaning it knows nothing about events beyond that point unless it uses tools like Bing search for live information.

Currently, ChatGPT’s cutoff is June 2024. This means if you ask about recent political changes or new tech inventions, it must search the web rather than rely on its training data.

5. ChatGPT hallucinates

AI hallucination is real, and ChatGPT is no exception. Sometimes it will generate information that sounds perfectly reasonable but is completely wrong.

This happens because it predicts what “should” come next in a sentence based on patterns, not on fact-checking.

6. ChatGPT uses calculator for math problems

Despite its sophisticated language abilities, ChatGPT doesn’t “do” math in its head. For complex calculations, it uses a built-in calculator alongside its reasoning skills.

This method, called chain-of-thought reasoning, allows it to break down problems step-by-step before arriving at the correct answer.

7. ChatGPT can be fine-tuned to learn from you

While ChatGPT doesn’t learn like a human, it can be fine-tuned to specialise in specific topics or styles based on additional training data.

For instance, businesses can customise it for customer service, medical advice, or legal support.

Even without official fine-tuning, ChatGPT can adapt to your personal communication style during a conversation.