AI image generators are trained on a large quantity of training data -- images that include human-generated labels over parts of each image. This process continues with additional filtering to ensure better representation and to weed out inappropriate images. Finally, through machine learning, the program learns to identify images without the need for labels. (Source: OpenAI)
Note that most AI image generators are trained to not provide graphic or violent images and most will not create images containing famous people. Nor will it create images in the style of a painter who produced works in the past ~100 years. For instance, it will generate images created in the style of Van Gogh, as seen here, but not in the style of Monet, whose final painting was completed in 1920.
Image Credit: Image created on Bing Image Creator on 15 May 2023 with the prompt: "a field of poppies on a partly cloudy day painted in the style of Van Gogh."
Find additional AI examples and mini-tutorials here.
Image created with Bing Image Creator/DALL-E 3 on 8 October 2023 with the prompt: Create a fuzzy phone picture of a fawn in the forest. It looks into the camera before running away.
This project was made possible in part by the Institute of Museum and Library Services and the New Hampshire State Library.