Blog Post #4

What ethical concerns do you have (or not have) about the use of some of these [AI] tools?

There are some legitimate ethical concerns regarding use of some of these AI tools. There are some tools such as Quillbot and other types of text generation tools that can be used and are even advertised to students to use to write their papers and assignments for them. While these advertisements are unlikely to directly promote academic violations, should students actually submit just the work that is generated by these tools with no other type of proper citations or accreditations that can be a taken as plagiarism.

There are also the image generation AI tools that are available to many. Many of these tools will be trained on images and art available on the internet and art sharing websites like DeviantArt without the consent of the creators. This poses an ethical concern of people using these tools and training them using certain artists works and putting it out as their own original work. This discredits and ignores all the time and skill that people put into their creations and for some it takes away from their livelihoods.

How accurate or successful were the learning objects you created using the AI tools?

I utilized the two image generation AI tools Stable Diffusion and DALL-E in my testing. I was interested to see how similar or different the two AI tools would respond to the same prompt. I did this test with a few different prompts. Most of my prompts were taking modern objects or characters and showing them in a traditional art style. Below are the images generated with the prompt “Hatsune Miku in the style of Picasso”. Some other prompts I had were “modern gaming PC in the style of Monet” and “squishmallows in art nouveau”. These other prompts produced images that weren’t as coherent as the images presented below. They may have vaguely looked like what the prompt was requesting, but only with the context of the prompt. With the images below I found that the Stable Diffusion images had textures that mimicked traditional paintings compared to DALL-E’s images having purely flat colours. Overall it seems like Stable Diffusion was more successful with my prompts and provided more detailed images, likely due to having a larger database to draw from and taking longer to generate the images. Although I have to say that the final image that DALL-E provided might be my favourite.