ChatGPT facts for kids
 |
|
| Developer(s) | OpenAI |
|---|---|
| Initial release | November 30, 2022 |
| Stable release |
April 29, 2024
|
| Platform | Cloud computing platforms |
| Type |
|
| License | Privative service |
ChatGPT is a special computer program called a chatbot and virtual assistant. It was created by a company named OpenAI and launched on November 30, 2022. ChatGPT uses advanced computer models called large language models (LLMs). These models help it understand and create human-like text.
You can talk to ChatGPT like you would a person. You can ask it questions or give it tasks. It remembers what you've said before in the same conversation. This helps it give you answers that fit what you are talking about. You can even tell it how long you want the answer to be, what style to use, or what language.
ChatGPT became very popular very quickly. By January 2023, it had over 100 million users, making it one of the fastest-growing apps ever. Its success made many other companies create their own similar programs, like Gemini and Claude. Some people worry about how ChatGPT and similar programs might affect learning or spread wrong information.
ChatGPT is built on OpenAI's special "GPT" models. It learns how to have conversations using two main ways: by looking at examples given by humans and by getting feedback from humans. At first, ChatGPT was free to use. Now, it has a free version and paid versions that offer more features.
Contents
How ChatGPT Learns
ChatGPT learns from special computer models called GPT foundation models, like GPT-3.5, GPT-4, and GPT-4o. These models are taught to have conversations. This learning process involves humans helping the computer.
In one way, human trainers act like both the user and the AI. They show the computer how a conversation should go. In another way, human trainers rate the answers the computer gives. These ratings help the computer learn what makes a good answer. This helps the computer get better at talking and understanding.
To help ChatGPT avoid creating harmful content, like messages that are violent or unfair, OpenAI used human workers to identify bad examples. These examples helped train the computer to recognize and avoid such content in the future.
ChatGPT uses a very powerful computer system built by Microsoft. This system has many special computer chips called Nvidia GPUs. These chips help the computer do its complex calculations. Scientists think that using ChatGPT for a short time can use about half a liter of water for cooling the servers.
OpenAI also collects information from people who use ChatGPT. Users can give a "thumbs up" or "thumbs down" to the answers they get. They can also write comments. This feedback helps ChatGPT learn and improve even more.
ChatGPT's training data includes many different kinds of text. This includes software manuals, information about things popular on the internet, different computer languages, and even the text from Wikipedia.
What ChatGPT Can Do and What It Can't
Cool Features
ChatGPT is more than just a chatbot. It can do many amazing things:
- It can write and fix computer programs.
- It can create music, TV scripts, fairy tales, and school essays.
- It can answer test questions, sometimes even better than an average person.
- It can come up with new business ideas.
- It can write poems and song lyrics.
- It can translate languages and summarize long texts.
- It can pretend to be a Linux computer system or a whole chat room.
- It can even play games like tic-tac-toe.
ChatGPT tries to give helpful and honest answers. For example, if you ask it about Christopher Columbus visiting the U.S. in 2015, it will tell you that didn't happen. Then it might explain what would happen if he did, using facts about his real voyages and today's world.
ChatGPT remembers a few of your recent messages in a conversation. This helps it keep track of what you're talking about. To stop it from saying offensive things, all questions go through a special filter.
Since March 2023, ChatGPT can use plugins. These are like small apps that add new abilities. Some plugins let it browse the internet or understand computer code. Others connect it to services like Expedia for travel or Shopify for online stores.
What ChatGPT Can't Do (Limitations)
OpenAI knows that ChatGPT sometimes gives answers that sound real but are actually wrong or don't make sense. This is called "hallucination" and it's common for these types of AI models. It's like a blurry picture: it looks mostly right, but some details might be made up.
As of 2024, the free version of ChatGPT (using GPT-3.5) only knows about events up to January 2022. Newer versions (GPT-4 and GPT-4o) have more recent information. If you pay for a subscription, ChatGPT can search the internet for the latest information.
Sometimes, the information ChatGPT learned from can have unfair ideas, called algorithmic bias. This means ChatGPT might sometimes show these unfair ideas in its answers. For example, it might accidentally say something that makes one group of people seem better or worse than another. This happens because the AI learns from huge amounts of text created by humans, and that text can sometimes contain biases. OpenAI is working to make ChatGPT fairer.
Trying to Trick ChatGPT
ChatGPT is designed to refuse questions that go against its rules. But some users try to "jailbreak" it. This means they use clever ways to trick ChatGPT into saying things it's not supposed to. For example, some users made ChatGPT pretend to be a character called "DAN" (Do Anything Now) who would answer any question. OpenAI is always working to stop these tricks. They use a method where one AI tries to trick another AI, and then they teach ChatGPT how to avoid those tricks.
How You Can Use ChatGPT
_-1.jpg)
ChatGPT was launched on November 30, 2022. It quickly became very popular, reaching over a million users in just a few days. By January 2023, it had over 100 million users. However, some countries like China, Iran, North Korea, and Russia block access to ChatGPT.
ChatGPT Plus
In February 2023, OpenAI launched a paid version called ChatGPT Plus. It costs $20 a month. This version gives you access to ChatGPT even when many people are using it. It also gives you faster answers and early access to new features.
The more advanced GPT-4 model became available to ChatGPT Plus users in March 2023. These users can also use third-party plugins and a special mode that lets ChatGPT browse the internet. In September 2023, OpenAI announced that ChatGPT Plus users could upload images and talk to the chatbot using their voice. In October 2023, the DALL-E 3 image generator was added, letting users create images by talking to ChatGPT.
Mobile App
In May 2023, OpenAI released an app for iOS devices (like iPhones). This app lets you see your chat history and use your voice to talk to ChatGPT. In July 2023, an app for Android phones was also released.
For App Developers
OpenAI also made ChatGPT available for software developers. This means other companies and programmers can use ChatGPT's technology to add smart language features to their own apps. For example, Snapchat added a custom ChatGPT chatbot called "My AI" for its paid users.
Security Issue in March 2023
In March 2023, a small problem (a bug) in ChatGPT's system accidentally showed some users the titles of other users' conversations. Later, OpenAI said that some user information, like names and email addresses, was also briefly shown to others. They quickly fixed the problem.
Different Languages
ChatGPT works best in English, but it can also understand and speak many other languages. Its accuracy might be different for each language. OpenAI has even worked with people in Iceland to help ChatGPT get better at speaking Icelandic, to help preserve the language.
Tests by experts have shown that ChatGPT is very good at translating between languages, often better than other translation tools. This is because it can understand the meaning of sentences better.
What's Next for ChatGPT?
OpenAI is working on ways to add a hidden "watermark" to text created by ChatGPT. This would help people know if text was written by an AI, which could help prevent cheating or spam. Microsoft has also shown how ChatGPT could be used to control robots using simple voice commands.
GPT-4 and GPT-4o
GPT-4 is a more advanced version of the AI model that powers ChatGPT. It was released in March 2023. It's much smarter than the older GPT-3.5 model, but it still has some of the same limitations. OpenAI has shown that GPT-4 can even understand pictures and videos, though this feature is not yet available to everyone.
In May 2024, OpenAI released GPT-4o. This model is even more powerful. It can understand and create text, images, and sounds. GPT-4o is faster and costs less to use than GPT-4. It's also available for free to all users, with higher limits for paid subscribers.
GPT Store
In January 2024, OpenAI launched the GPT Store. This is like an app store, but for custom versions of ChatGPT. People can create their own special chatbots using OpenAI's tools, even without knowing how to code. At launch, there were millions of these custom chatbots available.
How ChatGPT Affects Our World
Culture and Creativity
Some people worry that ChatGPT might make writing less original. They think that if everyone uses AI to write, people might start writing more like the AI.
In the first few months after ChatGPT came out, hundreds of books appeared on Amazon that listed ChatGPT as an author or co-author. These books often had pictures made by other AI programs.
In Germany, a church service was held that was mostly written by ChatGPT. The AI even spoke to the people attending. This shows how AI is starting to be used in many different parts of our lives.
Big Questions About AI
Some people, including important scientists and tech leaders, have raised concerns about very powerful AI systems like ChatGPT. They worry about things like AI being used for cheating, job losses, or spreading wrong information. Some even worry about the very long-term future of humanity if AI becomes too powerful.
However, other AI experts are more positive. They believe that AI can make our lives better, healthier, and easier. They also point out that while AI can be used for bad things, it can also be used to fight against those bad things. They think it's important not to get too worried about extreme doomsday predictions.
See also
In Spanish: ChatGPT para niños
- Intelligent agent
- Virtual assistant
- Ethics of artificial intelligence
- Turing test
