One issue that arises from combing AI and image manipulation is the overall accuracy of the image. In the DeOldify image colorize for example, we don’t 100% know whether or not the colors that it chooses for the new image is true to life. In a way this inaccurate representation seems useless in the fact that it takes away from the real life historical accuracy. In my art history class we had a unit where we learned about the rise and commercialization of photos. Similarly there was an example of a black and white photo of an Iranian family that was spread around on postcards and later changed into a colorized image. Not only was the original image a staged scene of a family taken to satiate the curiosity of westerners but the new image also pushes a certain narrative by the colors they chose.
“Adding color does not show things as they were but recreates what is already a recreation – a photograph – in our own image, now with computer science’s seal of approval.” – Sonja Drimmer
This quote further illustrates why colorizing old photos is essentially useless, especially in historical/humanistic fields. In these types of fields the focus is taking whats already there and learning from it, not learning from a computer generated recreation. I also don’t completely understand why any field would use image colorization, there doesn’t seem to be any new benefits that you can gain from the colorized photos. What I mean by this is that any new information possibly gained from the new photos is doubtful in terms of accuracy.
I can see how colorizing photos might be useful in a non academic and non historical context. For example, colorizing old family photos in order to preserve memories and bring them to life.
“if you’re looking for an exact sequence of bits, you won’t find it; all you will ever get is an approximation” – Ted Chiang
AI like Chat GPT and DeOldify will never be 100% accurate and if you are to use AI you should always bear this in mind. This is why it’s probably not the best to use in research or academic settings. In class we talked about using AI to help with basic coding and some people were for it, some against. But even though coding AI is different in the sense that you don’t have as much of a chance at accidently plagiarizing something. It is still useful to at least know how to code in order to catch small mistakes that AI might make.
Issues with DeOldify:
Unfortunately as I was downloading DeOldify it did not allow me to download the final step. I tried doing what they suggested in the error message but it still failed. I think there is something in the code that my computer doesn’t allow but I don’t know how to change that.
