🎨 AI That Makes Pictures

5th Grade 🤖 Artificial Intelligence

Generative AI is artificial intelligence that creates new content — text, pictures, music or videos. Image generators take a description in words, like “a cat riding a bicycle in outer space,” and create a picture that didn’t exist before.

Free · US Letter · no sign-up
Google Classroom Save to Pinterest Email me

From recognizing to creating

For many years, most uses of AI were about recognizing things: Is there a cat in this picture? Is this message spam? Whose face is trying to unlock the phone? Generative AI goes the other way — instead of recognizing what’s in a picture, it makes a picture based on what you asked for.

To do that, the program is trained on a huge number of pictures, many of which come with a description in words. That’s how it learns connections between words and how things look: what a “sunset,” a “poodle” or “a watercolor painting” usually looks like.

From noise to a picture

One well-known way of making pictures works in a surprising way. During training, you take a real picture and add “noise” — random dots of color — step by step, until it looks like the static on an old TV screen. The program learns to go the opposite way: to remove a little noise at a time and bring the picture back.

When you ask for a new picture, it starts from completely random noise, and the program gradually “cleans” it, with the word description steering it toward where to end up. After a number of steps — sometimes dozens — a new picture appears. It’s a bit like a sculptor chipping away at a block of stone until a figure appears — only here, it’s all numbers and calculations.

Because the process starts from random noise, the exact same description can give a different picture every time.

Weird mistakes

Image generators can make very beautiful pictures, but they also make strange mistakes: sometimes a person has too many fingers, the words on a sign look like jumbled letters, or a shadow falls the wrong way. This happens because the program doesn’t understand the world the way we do — it creates something that looks like the pictures it has seen.

These mistakes are getting rarer, so you can’t rely on them alone to spot a fake picture. Ask questions: Who posted the picture? Do trustworthy news sites show it? Does it have details that don’t make sense? And if a picture looks too amazing to be true or makes you really angry — that’s a good reason to stop and check before you share it.

Questions of fairness and rights

Image generators learn from pictures that people — photographers, painters and illustrators — created. That sparks a debate: Is it okay to use artists’ work to train a program without asking them? Who owns a picture a program made? Different countries are holding legal discussions about this, and the answers are still being worked out.

Another problem is misuse: people can create fake pictures of real people, as if they did something they never did. That can do serious harm. So never create or spread fake pictures of friends, teachers or anyone else, and tell an adult if you come across a picture like that of you or someone you know.

Creating responsibly

An image generator can be a fun tool for creativity: imagining a made-up creature for a story, trying out different art styles or designing a greeting card. When you use one, it’s fair to say the picture was made with AI, especially in schoolwork.

And remember: the program helps, but the idea, the choices and the taste are yours. You can also mix the two — make a draft with a computer and then draw your own version by hand.

There are also tools that make music, voices and short videos using the same basic idea: learn from lots of examples and create something new that looks or sounds like them. The rules are the same — create responsibly, say that you used the tool, and never imitate a real person’s voice or face without their permission.

Key facts

  • Generative AI creates new content: text, pictures, music and videos.
  • Image generators are trained on lots of pictures that have descriptions in words.
  • A common method starts from random noise and gradually cleans it into a picture.
  • The same description can give a different picture each time.
  • Creating and spreading fake pictures of real people can do serious harm.

Quick quiz

  1. What does generative AI do?

  2. What is an image generator trained on?

  3. In the well-known method for making pictures, what do you start with?

  4. Why can the same description give different pictures?

  5. Why do image generators sometimes get the number of fingers wrong?

  6. What should you do when you see an amazing, surprising picture online?

  7. Which use of an image generator is harmful?

  8. What’s the fair thing to do when you turn in work with an AI-made picture?

💭 Think and talk about it

  1. If a program made a picture from your description, is it your creation, the program’s or the artists’ it learned from? Explain why.
  2. What three signs would you check to suspect a picture online isn’t real?
  3. How can you use an image generator to help your creativity instead of replacing it?

Questions kids ask

Is a picture made by AI real?

The picture itself exists, but what you see in it didn’t happen in real life. It was created by calculations, even if it looks like a photo.

Does an image generator copy existing pictures?

Usually it creates a new picture rather than copying one, but researchers have found that sometimes it reproduces pictures it saw in training almost exactly. It also learned from other people’s pictures and can imitate styles. That’s why there’s a debate about fairness to artists.

How can you tell if a picture was made by AI?

You can’t always know for sure. Check who posted it, look for the picture in trustworthy sources, look for details that don’t make sense, and ask an adult when you’re not sure.

🔎 Ask Thinkapedia another question