Module 2 Generative AI and Evaluating Content and Tools for Learning

Craiyon. A Bear Riding an Orca. Craiyon.com, 9 Oct. 2024.

The SAMR generated by ChatGPT:

SAMR Analysis of Craiyon (Generative AI Tool for Learning)

Substitution:
At the substitution level, Craiyon can be used to replace traditional image-creation tools such as hand-drawn illustrations or online image searches. Instead of asking students to search for or create images manually, the tool generates visuals from text prompts. The learning task itself doesn’t change but becomes more efficient.

Example: A student could use Craiyon to generate an image of a historical event instead of manually drawing or finding one online.

Augmentation:
At this stage, Craiyon enhances the process by adding functionalities that go beyond what manual tools offer. For instance, the ability to generate a wide variety of images in real-time provides instant feedback, enhancing creativity and flexibility in learning. Students can iterate quickly and explore different visual possibilities with ease.

Example: A student could input different descriptions of a character from a book and receive various visual interpretations, helping them think critically about the text.

Modification:
Generative AI tools like Craiyon start transforming the task at the modification level. Here, students can modify their creative process, using AI not only to generate images but also to experiment with visual storytelling. The tool allows students to prototype, compare, and contrast visual designs in ways that were previously impractical.

Example: In an art class, students can experiment with different artistic styles for their projects by generating several iterations of a concept through AI. This supports deeper engagement with concepts like symbolism or theme.

Redefinition:
At the redefinition level, the use of Craiyon can enable completely new tasks that were previously inconceivable. Generative AI allows students to engage in collaborative storytelling where text and visuals are continuously generated and iterated upon by both humans and AI. It also promotes the exploration of creativity and problem-solving at a scale and speed that weren’t possible before.

Example: Students could work together to design an entire visual world using Craiyon, incorporating real-time feedback from the AI. This task integrates art, design, creative writing, and collaboration in an entirely new form of learning.

My Reflection:
I think one of the most fascinating things about AI image generation and I have tinkered with a few but never Craiyon is that the LLM uses it’s knowledge of what something should look like based on it’s knowledge but it doesn’t know what is right and is wrong once it has generated it’s image. So the image itself is the hardest information for it to understand.

When I prompted ChatGPT again to analyze what was in the above image this was it’s response:

The image appears to depict an animated or stylized scene of a bear riding an orca (killer whale). The bear is sitting atop the orca as it jumps out of the water, and the bear seems to be holding something, possibly a small backpack or an object on its back. The water is splashing around them, and there’s a small orange crayon floating in the water, likely a reference to the Craiyon AI tool.

This scene seems playful and whimsical, combining elements of both wildlife and fantasy.

ChatGPT. “Description of Image: Bear Riding an Orca.” OpenAI, 9 Oct. 2024.

So as you can see ChatGPT completely missed that the Orca has 2 sets of fins and an extra dorsal fin because it can’t tell what a correct orca should look like only that it looks like an Orca.

So my Ethical concern with AI art as that itself is a feedback loop. The more AI art that is generated that is wrong. The more art out there that is wrong it uses to create more wrong AI art. Therefore it will eventually poison the sample supply to far beyond repair If AI art continues to spread across the globe.

But as of right now I think AI art is a great tool for learning. It is not always super correct but I have used it myself to generate art for slideshow projects because it is a great way to produce art for abstract thoughts in the moment. Like If I wanted to create a slideshow about orcas and bears it would be hard to find an image online that has both in them that does not have copyright laws protecting them.

That being said there is a grey area of AI art sometimes sampling copyright protected images but I imagine that will be sorted soon in the future by lawyers.

I think that in 2-3 years time AI art will either be so good it replaces artists or so bad it creates the negative feedback loop I mentioned and becomes really terrible. But only time will tell. Right now If you want to test my theory on the feedback loop google anything abstract like “Bear licking ice cream” or “Cat in car” and at least ONE of the top 5 results will be AI Art (It’s sometimes all of them). It’s like a plague on the internet but at the same time a gift.

4 Comments

  1. linkechen

    Hello Liam! This is a really well-written blog. In my studies this week, I also tried using Craiyon to generate images, and as you mentioned, AI-generated images seem to rely on its own knowledge base and the content you provide. However, the lack of timely updates to the knowledge base does lead to more and more inaccurate images, which seems to be a common issue with many AI tools. As for the content of the blog, I think it’s great. The only improvement I would suggest is in the formatting—perhaps using headings to distinguish different sections would make the structure clearer. Overall, it’s a fantastic blog!

  2. jinwonk

    Hello Liam, it’s very interesting! I blogged about ChatGPT, while you blogged about Craiyon, which I don’t know much about, and I found it very interesting. I learned more about Craiyon through your blog. I agree with your opinion that AI art is a good tool for learning. It’s because we can easily create images that we only think of abstractly. However, there are also clear problems with AI art, so there are areas where they need to continue to receive feedback and improve them. Great blog!

  3. ivypav

    Hi Liam!! I also performed a SAMR analysis of Craiyon, and I agree that its a great tool for visualizing abstract thoughts or creating images for unconventional/novel situations (which I find interesting as a linguistics student since the number of comprehensible English sentences we can generate is countably infinite, while our artistic capability may not be representative of this). I also found it very that you created interaction between the two AI tools you used by having ChatGPT describe the image Craiyon produced, I would never have thought to do that. Awesome post!

  4. jungjooyoon

    Hi Liam! Your post does a great job explaining the good and bad sides of AI art. I really liked how you talked about the feedback loop and how mistakes in AI art could lead to more mistakes in the future. It’s a clear and thoughtful look at both the benefits and issues with AI art!