Like a Copycat Kid
Imagine you have a friend who always says, "Boys play with trucks, girls play with dolls." You hear this again and again. Soon, you start to think that's how things should be, even if it's not true. AI models are like that copycat kid: they learn from what people say and do.
Learning from Examples
AI models read lots of books, messages, and stories, just like you read your favorite storybook every night. If most of those stories say boys are brave and girls are gentle, the AI starts to think that’s normal too. It might even start to copy those ideas when it talks or writes.
So, if an AI model says something like, "She's a nurse, he's a doctor," it's just repeating what it learned, like your friend who always says boys play with trucks and girls play with dolls.
That's how gender stereotypes and bias can become part of the way AI models think.
Examples
- An AI assistant suggests a boy should play with trucks and a girl should play with dolls based on the data it learned from old advertisements.
- A hiring tool prefers male candidates for leadership roles because it was trained using historical job applications that favored men.
- A language model uses words like 'aggressive' to describe women more often than men, showing how bias can be passed down through sentences.
Ask a question
See also
- How do AI image generators create such realistic art?
- How do AI image generators create realistic pictures?
- How do AI systems develop gender bias and stereotypes?
- What is Generative AI?
- How do deepfakes work, and can they be detected?