A generative AI model can use copyrighted material for training if it’s allowed by the copyright owner or if it follows certain rules.
Imagine you're building a toy robot that learns to draw pictures by looking at lots of drawings. If the robot copies a drawing exactly, like copying your favorite picture from a book, then the person who made that picture might not be happy, because they own the copyright of that picture.
But if the robot looks at many different drawings and uses bits of them to learn how to draw new pictures, it's more like learning from a big art class. In this case, the copyright owner might let it do that, or maybe even pay for it.
What happens if they don’t agree?
If the robot copies too much without asking, the person who owns the picture might say, “Stop copying!”, just like how you’d stop your friend from taking all your toys without asking. That could lead to a legal problem, like getting a letter or even having to pay for using the pictures.
So it’s about learning and sharing, sometimes with permission, sometimes not!
Examples
- A child uses a picture from a book to draw a new picture without asking the author.
- A baker uses a recipe from a cookbook to make a cake for a party.
- A student copies text from a textbook to write an essay.
Ask a question
See also
- How do Generative AI models learn to create new content?
- How do generative AI models create images and videos?
- How do generative AI models like ChatGPT create new content?
- How do generative AI image models create realistic pictures?
- How do AI deepfakes trick people so easily?