Even from a young age, in school, we learn the importance of primary sources. When creating historical analysis or forming an argument, it’s important to get at least some of our information directly from the source. Historical photos are primary sources. When they’re colorized, especially using AI, that’s no longer the case. However, this can be deceiving.
Using artificial intelligence to color old black and white photos can help to make the scenes within them feel more real. However, when the use of AI isn’t accurately attributed, it threatens to mislead the viewer, and can create falsehoods about the actual colors, historical background and artistic integrity of the photo.
We can see this falsehood in practice. For example, I used the AI to colorize an old photo of Nourse Hall from the Carleton Archives. While it was mostly accurate, the AI ultimately dulled down the color of the building, making the bricks appear dark brown instead of red and turning the roof gray. We know this is incorrect because we can see the building today, as well as some color photos of it from early in its time, but if the photo wasn’t labeled, we can see how someone can be misled.
Beyond just the concerns of historical falsehoods, AI generated images also have a plagiarism problem. This has been discussed widely over the past few months, but no conversation about AI and art is fair without mentioning how artificial intelligence essentially is regurgitating artwork from real people. Specifically in the field of analyzing photos, as Sonja Drimmer points out, there is already a group of experts who can analyze old pieces of art and explain their significance, methods, as well as describe what colors were likely in these photos. AI is just doing a less complete version of this task.
[Art historians] explore how works of art shaped the worlds in which they were made and would go on to influence future generations. A computer algorithm cannot perform these functions. Drimmer, Sonja. “How AI Is Hijacking Art History.” The Conversation, The Conversation, 1 Nov. 2021, theconversation.com/how-ai-is-hijacking-art-history-170691.
However, this is a wider problem than just colorized photos. As Ted Chiang writes, AI generated images and writing lack creativity, originality and complexity, so when we rely on AI to produce art and literature that would otherwise be made by humans, we’re dulling the world.
Obviously, no one can speak for all writers, but let me make the argument that starting with a blurry copy of unoriginal work isn’t a good way to create original work.
Chiang, Ted. “ChatGPT Is A Blurry JPEG of the Web.” The New Yorker, 9 Feb. 2023, https://www.newyorker.com/tech/annals-of-technology/chatgpt-is-a-blurry-jpeg-of-the-web. Accessed 24 Jan. 2024.
2 thoughts on “AI and Images”
Leave a Reply
This site uses Akismet to reduce spam. Learn how your comment data is processed.
I agree with you that AI may have accuracy issues in colorizing. I also noticed some deviations when using DeOldify, which can indeed lead to misinterpretation of history. Additionally, I’m concerned that this could bring up the issue of misleading advertising. While this technology is beneficial to us in many ways, I think we should limit its use in certain specific areas, especially in areas such as education and historical archives where maintaining the original authenticity of images is undoubtedly more important than pursuing artistry.
I think the image you created and displayed was a really good example of AI falsifying the colors of an image. If I hadn’t gone to school here I would think that the building was brown and grey. I agree that the issue of artificial intelligence plagiarizing information has become an increasingly discussed topic. You used the two readings really well to support your argument.