For the AI colorization lab I decided to use a picture of the Chapel interior. There was not an exact reason for this decsision but I thought it would be a good test for the AI. If you would like to look at the new photo in better detail go here: https://dgah.sites.carleton.edu/digtialobjects/admin/items/show/132. Below is the original picture drawn from the Carleton Archive with the AI colorized image below as well.
Distant photos can be brought into a more modern perspective. In general, I believe this can be beneficial in better depicting past moments in time. However, I don’t think it should be done in every case. The quote below captures my position well:
“The humanities will lose what makes them vital. The field’s sensitivity to historical particularity and cultural difference makes the application of the same code to widely diverse artifacts utterly illogical.”
AI-Generated Art and Historical Accuracy
In the context of humanities research, it is essential to interpret evidence and sources within the time and culture in which they were created. It is the role of humanities researchers to analyze these findings and connect them to the present. While colorizing an image allows us to view it through a contemporary perspective, it comes at the cost of distorting, generalizing, and ultimately losing valuable historical data.
“The more that text generated by large language models gets published on the Web, the more the Web becomes a blurrier version of itself.”
Xerox Photocopiers and AI Hallucinations
Just as AI-generated text risks turning the internet into a less accurate version of itself, the repeated use of AI tools on historical photos risks transforming the past into a muddled reflection of the present. AI does not “know” the true colors of historical images, it merely predicts them based on training data. As a result, it generates an output that aligns with its dataset rather than historical truth. The more we rely on these tools, the more we risk replacing the original integrity of historical photos with whatever data a company has chosen to feed its AI models.
For the humanities, the ability to analyze and interpret sources directly from their original context is one of the most critical skills in the field. As we offload these tasks to AI models, we become increasingly dependent on the companies that develop and control these tools. This also means placing significant trust in these companies to produce results that are reliable and suitable for scholarly interpretation. In my opinion, this is a risky path. AI is simply not advanced enough to handle sensitive tasks like colorization without compromising historical accuracy. The human element in the humanities is what makes it so valuable, and that should not be replaced by automated predictions.
2 thoughts on “Lab Assignment – Week 4”
Leave a Reply to Jennifer Cancel reply
This site uses Akismet to reduce spam. Learn how your comment data is processed.
Hi there, I really like how you choose to restore the chapel interior! Its truly quite remarkable how is has barely changed since then. It’s interesting to see how the light from the colored windows are shown through the color restoration… They’re like little light halos around the window which knowing how the chapel looks like is inaccurate. You’ve made a firm stance on AI’s inability to accurately handle sensitive tasks. What if we were to just use AI to create a general image/outline of whats possible? Would you say its ok to use AI if we had clear and well enforced laws about watermarking/ labeling AI materials?
I agree with the sentiment that ai can’t be relied on in the field of dh. While it most definitely is a useful tool, ai does not have the ability to break down historical content as humans do. It is interesting how the ai chose to colorize your photo, it seems to have missed the subtle variations in color and instead chose to use the same hue of brown for everything.