Module 2: Generative AI and Evaluating Content and Tools for Learning

AI generative image

“A jumping cat” prompt, Stable Diffusion Online, 11 Oct.2024, stablediffusionweb.com/app/image-generator/.

“A blackhole raven” prompt, Stable Diffusion Online, 11 Oct.2024, stablediffusionweb.com/app/image-generator/.

When given the option to generate anything using an Generative AI tool, I went with an image generator tool, Stable Diffusion. It was my first time working with this tool but thankfully the interface was very intuitive and user-friendly. As for what to generate, I wanted to work with a simple and an abstract image prompt. I followed this with description prompts for the generated images in ChatGPT.

The prompts I settled on and their description are as follows:-

1. A jumping cat (simple):
   The image shows a cat mid-jump, appearing to be airborne with all four paws off the ground. The cat has a focused expression, with its eyes wide and ears slightly back, as if it’s leaping towards something. Its fur is a mix of white and tabby stripes, and the background appears to be indoors with tiled flooring and a neutral-colored wall.
   
2. A blackhole raven (abstract):
   The image features a black raven in a surreal, cosmic environment. The raven is perched in front of a swirling, dark vortex or portal-like structure, with space and stars in the background. The lighting casts a soft glow from behind the raven, emphasizing the bird’s glossy feathers and creating a mysterious, otherworldly atmosphere. The overall mood is dark and ethereal, blending natural elements with cosmic and abstract designs.
   
   “Give me a short description on the images attached.”, prompt, ChatGpt, version 4.0, OpenAI, 11 Oct. 2024, chatgpt.com/

Reflections (Image)

When regarding image accuracy, I found the simple prompt to generate a very accurate image of a jumping cat with not much noise or deviations compared to my personal idea. As for the blackhole raven, it did differ quite a bit from my idea with it just being a raven and a black circle in the same image. However, that was to be expected since it was a very abstract or fantastical term. From this it can be seen that the LLM (Stable diffusion) has a knowledge bank that is heavily cluttered with existing ideas (bland and formulaic) from the internet but lacks the capacity of creativity.

Regarding description accuracy, I found the description for the cat to be very accurate if not excessively informative since I myself was only described it as “a jumping cat”. Whereas the blackhole raven description didn’t use the term blackhole at all. I believe this is due to the simple black circle with white interiors depicted in the image which can almost be described as a cosmic phenomenon.

AI generative text

Using ChatGPT to create a SAMR analysis on use of a generative AI tool (Perplexity) for learning

I have written down the analysis provided by ChatGPT below:

———————————————————————————————–

To create a SAMR (Substitution, Augmentation, Modification, and Redefinition) analysis on the use of Perplexity, a generative AI tool, for learning, we can break down the tool’s application across each level of the SAMR model. SAMR is used to analyze how technology integrates into learning and can transform educational practices.

Here’s how we can analyze Perplexity in this framework:

Substitution

At this level, technology acts as a direct substitute for traditional tools without any functional improvements.

• Traditional Method: Students would typically use textbooks, notes, or perform manual searches in libraries or on the web to answer questions.
• Use of Perplexity: Instead of searching manually or referring to books, students ask Perplexity directly to get answers.
• Example: A student asks Perplexity a factual question like “What is the capital of France?” and gets the same answer they would have obtained by searching online or looking in a textbook.
• Impact: Perplexity simply replaces the conventional information retrieval method (manual search) without adding any functional benefit beyond speed.

Augmentation

At this level, technology provides direct tool substitution with some functional improvement.

• Traditional Method: Students search multiple sources, evaluate each one, and synthesize information.
• Use of Perplexity: Perplexity not only provides an answer, but also generates concise summaries from various sources, improving the efficiency and depth of information retrieval.
• Example: A student asks Perplexity for a summary of a complex concept like “quantum computing” and receives a coherent, synthesized explanation that includes definitions and applications, saving time and effort in synthesis.
• Impact: This improves learning by speeding up the process of obtaining reliable, synthesized information and offers a clearer understanding of complex topics.

Modification

Here, technology significantly redesigns the task, enabling learners to achieve outcomes that were previously impossible or difficult.

• Traditional Method: Students manually read articles and synthesize multiple viewpoints, which can take considerable time.
• Use of Perplexity: Students can interact with Perplexity to ask follow-up questions, request clarification, and even seek different perspectives in real time.
• Example: A student working on a research project asks Perplexity for different viewpoints on climate change, and then asks it to explain the scientific consensus vs. dissenting opinions, refining their understanding in an interactive and dynamic way.
• Impact: The tool modifies the learning experience by providing interactive and iterative learning, where students can continuously adapt and deepen their research based on evolving needs and follow-up questions.

Redefinition

At this level, technology allows for the creation of new tasks that were previously inconceivable.

• Traditional Method: Learning is largely limited to searching for information and writing reports or essays on a specific topic.
• Use of Perplexity: Perplexity can be used to create entirely new learning experiences, such as personalized learning paths, creating adaptive content that evolves based on the student’s responses, or generating simulations to explain abstract concepts.
• Example: A student asks Perplexity to simulate different economic models and how changes in policy could affect real-world markets. They are not just learning passively but are engaging in predictive learning models.
• Impact: This transforms learning by enabling students to engage in real-time simulations, personalized content generation, and the creation of unique, interactive learning scenarios that were not possible without the AI tool.

This SAMR analysis shows how Perplexity can progress from simply replacing traditional tools to transforming the learning process entirely.

———————————————————————————————–

“help me create a SAMR analysis on the use of a generative AI tool Perplexity for learning” prompt, ChatGpt, version 4.0, OpenAI, 11 Oct. 2024, chatgpt.com/

Reflections (Text)

Regarding the things that are the hardest to assess from the above generated information would be regarding limits of Perplexity’s features and it’s impact on long-term learning. While it provides information on a few of it’s features clearly, it shouldn’t be the limit. Similarly, it only provides it’s immediate effect or transformations on learning, while having limited insight on long-term learning. The model tells us that this tool has the ability to enhance (Substitution & Augmentation) and transform (Modification & Redefination) learning. As for what it missed, it would be regarding ethical or bias concerns which depends on the knowledge bank used by the tool. It also doesnt give any information on it’s generation and information accuracy.

I have been using AI tools, especially ChatGPT for quite a while now. With enough frequency to know its strength and weaknesses. I have seen it’s usefullness and higher accuracy mostly when I prompt shorter questions or for shorter sections of code. I mostly use it when I dont understand some terms or I want it to go through my code to check what I did wrong. But I also know that it is really bad when it comes to heavy mathmatical calculations. For example, when working with boolean algebra simplification problems, it sometimes downright gives a wrong answer when the equation is to long or complicated.

I think that AI technology has especially bloomed over the past few years and we have seen constant development around it. So, while 2-3 years seems short when we talk about innovation, I believe that these tools will only become more comprehensive in their generation, be it from updating of their knowledge bank or stronger machine learning technology. I also know the growing fear of AI tools replacing human labour, but I believe that on a more positive note that AI tools will focus more on enhancing human creativity rather than replacing it.