During 2022, three tools brought AI image creation to the public. DALL·E 2 from OpenAI was announced in April, Midjourney opened its public beta in July, and Stable Diffusion was released as open-source software on 22 August.
How do they work?
These systems learn from huge collections of pictures paired with text descriptions. They discover patterns, for example what "a red bicycle against a blue wall" tends to look like, and then generate new images that match a written prompt. They do not copy-paste existing pictures, but they are shaped by the examples they have seen.
Why it is exciting
Anyone can now sketch an idea in seconds, test different looks for a poster or illustrate a story without drawing skills. Teachers can create classroom visuals quickly.
Why people are worried
- Artists' work was used to train these systems, often without permission or payment, and the courts and lawmakers are still working out what that means.
- Fake images can look convincing, so we must teach children to ask where a picture came from.
- Bias. The pictures reflect the data behind them, and that data does not represent everyone equally.
Our approach
On this site, the photographs of our classes and events are always real ones. We use AI tools thoughtfully, and we teach learners to question what they see.