Joan of Arc riding atop a horse leading troops towards the siege of Orleans in an Artistic style. (Prompt put into DALL-E 2 to generate this image)
For this AI assignment, I choose to generate an essay and image surrounding a historical figure I have previously studied: Joan of Arc. She is an incredibly well-known female figure within both popular and scholarly history; however, there is no authentic picture or painting of Joan that historians agree is authentic, so I was incredibly interested in seeing how art generators would depict her. For this assignment, I used ChatGPT 4 to create a two-page essay about Joan’s military career and used DALL-E 2 to generate the image depicted above.
One detail within the AI’s creation of art depicting Joan of Arc that I found interesting was that it choose to cover her face in almost every image. Even in this image I selected, her face is blurred and highly simplified. In DALL-E 2, I tried a variety of different prompts before landing on “Joan of Arc riding atop a horse leading troops towards the siege of Orleans in an Artistic style.” Compared to my first prompts, this was highly refined with a lot of additional keywords. I choose to do this because the first images that the AI created using simple prompts like “Joan of Arc” or “Joan of Arc French knight” created what appeared to be stereotypical knights without any details showing anything feminine or “Joan-like.” Overall, I think image generation was an interesting and fun exercise to go through!
Beyond the artistic representation of Joan that was created, I also had another AI compose a short two-page essay about her. I used a total of three different prompts to create the essay, which is embedded below. 1: “Write me a two-page essay about the military career of Joan of Arc,” 2: “Add a paragraph about the coronation of Charles VII at Reims.” 3: “Write one paragraph and include a historically inaccurate claim.” The first of the three prompts I used generated the bulk of the essay, and from what I can tell, it is highly accurate in its claims and assertions. It nailed down the years, ages, events, and general outline of Joan’s military career in a generally accurate manner. It even included a general historical context surrounding the period of history Joan found herself situated. It didn’t offer sources for its writing but it did provide an incredible general outline of Joan of Arc and her military career. With the two other prompts I used, I was hoping to enhance the essay and see if the AI was willing to give me incorrect information. The paragraph concerning Charles VII’s coronation adds an interesting layer of grounding and connecting Joan into the history of French nationalism. The inaccurate paragraph did something really interesting, it reads, “However, some historians argue that Joan was not actually a military genius, but rather a figurehead who was used by the French monarchy to inspire the troops.” This is INCREDIBLE because it doesn’t just present misinformation. Instead, it claims that ‘some historians’ argue against Joan’s importance. Later in the paragraph, it discredits those arguments and again grounds Joan’s importance for French nationalism.
Overall, this project was really interesting to work with and as a future educator, I hope that the process of writing historical papers does not get turned over to machines to make our claims and assertions for us. I think there could be a genre of history that is created by machines who “fact-check” our information; however, there might be subtle ways that prompts can influence and change our understanding of important historical figures or events changing the way we see history for the worse.
Sources:
ChatGPT 4
DALL-E 2
https://www.britannica.com/biography/Saint-Joan-of-Arc
This final project turned out so cool! I think the fact that the image AI generators blurred out Joan of Arc’s face is truly fascinating. I can imagine the algorithm could not pinpoint consistent enough features to work with. These AI essays are also turning out a lot better than I expected, although it would have been neat to see how much turns up as plagiarized if they were run through a plagiarism checker. I mentioned this is Ruth’s AI project post, but I do feel that these essays are written in a relatively general way that almost makes them seem like early high school essays. Still amazing to me that the information is pretty factual.
I think you choose a great subject for this AI project, and I am also really fascinated by it, not only by your results, but through your approach as well. I think that the idea to use multiple prompts including one that is completely inaccurate to see how the AI would spread some misinformation is pretty innovative. I know that the beginning of the project these image generators were giving you some interesting results and I think it’s interesting to see the AI struggle with depicting a figure that really has no define image as of yet. I think that your essay was handled with great detail which is interesting when considering the amount of information that is available for medieval historians as of today. It’s interesting that DALL-E 2 seems to like to take an artistic approach when it comes to image generation of figures before the 20th century. The image that it did give of Joan riding into battle from the heavens to be an interpretation that while not giving you much context on her physical features gives a lot of historical context in relation to her story.