Finding Images

Artificial Intelligence, "Leonardo da Vinci drawing style" an image created by the Midjourney AI art tool. Search results in Google Lens.

Artificial intelligence technologies for image creation and analysis are making a huge impact on art, design, and the study of art history. What does this mean for academic research and for academic content creation? When using images in your scholarly work remember these key points:

Authenticity: When you use an image in your research, you must know and cite the source of that image. When using a general search engine, like Google images, always fact check your images and find the original source. These tools can help:

• TinEye Reverse Image Search. Submit an image to TinEye to find out where it came from, how it is being used, if modified versions of the image exist, or to find higher resolution versions.

• Google Lens. Google Lens lets you search by image. Using a camera or an image, Google Lens can try to find the original source of an image, or discover visually similar images and related content, gathering results from all over the internet. 

• Content Authenticity Initiative Verify. Verifies digital provenance and inspects Content Credentials in detail. 

Pro tip: If these reverse search engines do not find a source, or any similar images, that can be a red flag that they are AI generated.

• Kashtanova Decision. US Copyright Office - Zarya of the Dawn Letter. February 2023 (PDF.7.3 MB). In its first decision (February 2023) on AI images, the United States Copyright Office granted limited copyright protection to the graphic novel "Zarya of the Dawn", which was created using AI generated images. While the Copyright Office approved protection for the text and for the arrangement of the images, it denied protection for the individual images

• Bloomberg Law: AI-generated Art Lacks Copyright Protection, D.C. Court Says. (August 2023). A Federal Court recently upheld the (February 2023) decision that AI-generated images cannot be copyrighted, stating that, "Human authorship is the bedrock of copyright law."

• Is AI generating an 'averaged', one-sided view of Art History? (Art Newspaper, 31(358), 2023. Artists are getting creative to counter visual language being skewed by image-generating apps that average out scraped photos and social media files into 'mean images.'

• Is AI Art Stealing from Artists? Kyle Chayka (New Yorker February 10, 2023). According to the layer behind a new class-action suit, every image that a generative tool produces "is an infringing, derivative work."

• AI Art: How artists are using and confronting machine learning: How to see like a machine. (The Museum of Modern Art, 2023, streaming video 14 min 56 sec.). 

• Adobe Firefly. Adobe firefly creates image based on text prompts. Firefly is trained on the Adobe stock image library (with creator permission) and public domain images. This tool also adds content credentials to each generated image, making the source metadata of the image transparent. Firefly is free to download as part of the Adobe Creative Cloud suite of tools.

• DreamStudio AI:  A web-based suite of generative media tools - including image and video generation tools. 

• Midjourney. Midjourney is one of the most popular AI image generators. Access is through the Discord app. The dataset(s) used to train this tool is not transparent, however Midjourney has been known to use the LAION dataset in the past. The LAION dataset uses copyrighted material.

• Microsoft Designer: CoPilot image creator. A free text-to-image creator from Microsoft.

Glaze. The Glaze project is a free tool developed at the University of Chicago to protect artists from having their work scraped and used by AI image generators. Glaze adds hidden (pixel) data to digital works and online images to 'confuse' AI scraping programs. 

Nightshade. Developed by the University of Chicago, Nightshade disrupts AI image datasets by adding 'poison pill' data that leads the AI image generator to misidentify images and concepts. 

Have I Been Trained. This online tool allows anyone to upload an image to discover if their work or personal photographs have been used in AI training datasets.