We were given an assignment this week from Middlebury College’s Demystifying AI course. According to Professor Bond, this assignment will require us to use AI to make images, and it will definitely cause us to consider the repercussions of doing so. For this exercise, I used DALL-E 3 from ChatGPT4 and Stable Diffusion, two AI picture generators. I instructed these two AI image generators to create an image of Elvis Presley first. The result is displayed above. I believe that both image generators really hit it out of the park with this one, emphasizing his attractive features, his well-known hair, and even dressing him in authentic Elvis attire. Stable Diffusion produced the image on the right, whereas DALL-E 3 produced the image on the left. Since Elvis Presley is one of the most well-known singers, it is apparent that both AI images featured a microphone and a guitar that fit him. To be honest, I believe Stable Diffusion Elvis is a little closer to the real Elvis, but that’s just me being picky. Overall, both AI picture generators did a fantastic job of adhering to the prompt and covered all the bases for this first prompt.
I instructed both of the AI image generators to create an image of Elvis Presley performing at a concert because the following part of the assignment requires us to add some flair to our prior suggestion. These images are displayed above; DALL-E 3 produced the picture on the left, while Stable Diffusion produced the picture on the right. While the Stable Diffusion image appears to be an older and fatter version of Elvis, the DALL-E image appears to be a young, confident, and up-and-coming Elvis. These two images appear to have slightly distinct thought processes. Still, both pictures had spotlights, fans, and Elvis dressed like a star for a concert.Again, I believe the Stable Diffusion Elvis to be a little more realistic looking, but the DALL-E handled my prompt quite well. All in all, these AI image creators did another excellent job.
In order to complete this assignment, we had to design an AI image of something unusual that we couldn’t locate online, in accordance with our initial request. I requested an image of Elvis Presley dressed like a mermaid from both AI image producers. The two outputs are displayed above; DALL-E 3 produced the picture on the left, and Stable Diffusion produced the picture on the right. Elvis submerged as a mermaid, complete with all of the added elements, is what really made the DALL-E 3 generator shine. Although the steady diffusion image fulfilled my need, I feel that there are too few features in the image.It’s also frustrating that you can’t see Elvis’ entire mermaid tail. However, I believe both of the AI image creators fulfilled the assignment I assigned them.
We must consider everything that the AI picture generators have produced as the assignment’s last phase. The Stable Diffusion image of Elvis as a mermaid was the worst AI image I believed to have been produced, even though none of them were very awful. As I mentioned in my note beneath the image, I believe it is merely devoid of detail. Elvis’s torso with the mermaid tail is partially visible, but it ends abruptly. Though it’s missing several details that I requested to see when I requested this to generate, I still think the photo is nice. The DALL-E 3 picture of Elvis dressed like a mermaid was the one I found to be the greatest. With this one, the AI really showed off its creativity by dressing Elvis normally but adding a pink/blue mermaid tail. He is surrounded by a lot of seal life, which enhances the picture’s ambiance. In the final section of our reflection, we are asked to assess the extent to which we believe AI image generation poses a threat to artists. Because I think there is value in a genuine image, picture, or portrait, I don’t think AI image generation poses a serious danger to artists. While I think AI-generated images are visually appealing, it is evident that they are primarily artificial. I think artists will always have a place in this world because authenticity is what people love most.
You can find the assignment here, Middlebury Assignment.