What tools did you find useful in your explorations this week and how did you use them? Which ones were not useful?

In this scenario, I consider generating images with AI tools. As a newly developing online tool, Stable Diffusion is based on the diffusion technique to provide text-image services. See Figure 1.

Figure 1

While I browsed the Stable Diffusion website, I found the operating interface is simplism. Generally, users only need to type a few words they want AI to generate and choose aspect ratio. Eventually, users will get the image they want as click the “Generate” button. Furthermore, Stable Diffusion offers multiple image styles allowing users to select. However, while I was going through the AI tool, some flaws caught my attention.  Stable Diffusion may not contain certain accuracy.

Accuracy is extremely important for me. For instance, I typed “blue sky, three suns, desert, and several mansions” in text box, it only generated “blue sky, desert, and several mansions,” which lack the part of “three suns.” Figure 2.

Fig. 2. “Blue sky, three suns, desert, several mansions” prompt, Stable Diffusion, version 2, OpenCLIP, 17 Mar. 2024, https://stablediffusionweb.com.

According to figure 2, I am eager to know if I type more words in the text box, it will generate an image based on the original picture or create a new one. The result is evidence that the Stable Diffusion will generate a new image to replace the old one. This time I found more misunderstandings AI made. I was intend to create an image have blue sky, three suns, desert and Chinese architecture, and then I got this image.  Figure 3. Which has lost one subject “three suns.” These errors may not let outcome followed by people’s willing. Meanwhile, Stable Diffusion has a limitation that does not allow users to generate ten images per day, which could reduce effective of people’s work. If user want to use additional functions, they have to upgrade to premium at the first. Hence, the users’ experience will be affected by those limitations.

Fig. 3. “Blue sky, three suns, desert, Chinese architecture” prompt, Stable Diffusion, version 2, OpenCLIP, 17 Mar. 2024, https://stablediffusionweb.com.

Where do you think these tools will be in their evolution in 2-3 years’ time?

As AI models, I believe it will continue to learn and evolve, I can expect a significant improvement in accuracy. This means the images generated will more closely match the descriptions provided. Future text-image tools may be able to handle more complex and detailed descriptions, creating images that are more intricate and nuanced.

SECTIONS analysis

  • S tudents: Users of Stable Diffusion might be artists, painting learners, or AI enthusiasts who need a reliable tool for generating stable and consistent AI outputs. Stable Diffusion could support diverse user needs by providing a wide range of functionalities and customization options.
  • E ase of use: Stable Diffusion could be designed with a user-friendly interface that makes it easy for users to input their requirements and understand the outputs.
  • C osts: Stable Diffusion offers a simple operation way for users, which is easy to learn for newer and does not require learning for a long time. However, there are useful functions that need to be upgraded plan.
  • T eaching functions: Stable Diffusion could support a variety of strategies by providing diverse functionalities, and it also provides a range of space customization for users to explore new features.
  • I nteraction: Stable Diffusion could support interaction by allowing users to easily share and discuss outputs. It also provides features for users to give feedback or make suggestions for improvement by email.
  • O rganisational issues: My institution give us as much as room to explore AI tools , which could easily manage students’ workflow, connect with newly developing methods of education and address problems by more comprehensive thinking.
  • N etworking: Stable Diffusion cloud facilitate collaboration and communication among users by group of artists or AI enthusiasts. It might also support integration with other systems or platforms. 
  • S ecurity and privacy: Stable Diffusion should comply with relevant laws and regulations regarding data security and privacy. University of Victoria has published its policy about AI guidelines, which is maximum protect students while using AI.