Today, I will be taking you guys on this journey of choosing a digital literacies pathway with me. Honestly, I had no idea what that meant going into this assignment so, I didn’t have anything to expect. Now, having finished it, I can honestly say it was a very fun assignment. Probably one of the best we had.
First, I started by doing my digital confidence profile. It’s basically to let me know my level of confidence and comfortability in my digital skills. It’s clear that I believe I’m quite confident in my digital skills as I spent so much time on my computer for different purposes. I faced so many problems that I solved on my own which honestly helped me become this comfortable.
Secondly, I read Dr. Maha Bali’s article on Knowing the Difference Between Digital Skills and Digital Literacies, and Teaching Both. It was a rather short but very informative article. Honestly if someone asked me before what’s the difference between digital skills and literacies I wouldn’t have a good answer like I do now. Dr Bali says “Digital skills focus on what and how. Digital literacy focuses on why, when, who, and for whom.” She basically states that digital literacies are not about the technical side but, it has more depth to it. It’s about respecting other people’s work, protecting ourselves, and, most importantly, understanding the reason behind all this. To understand it better, she gives the example of using images in presentations or articles. Digital skills means helping people simply by downloading images and integrating them into their work. On the other hand, digital literacies focus more on choosing appropriate images, citing, recognizing copyrights, and possibly using alternative text for people with visual disabilities. It’s such a huge difference between them and we really need to be more knowledgeable in this aspect as it could become dangerous for one that isn’t familiar with the rules and regulations. She also mentions that digital literacies also means the judging and validating of the stuff we read or write and why we are doing it.
Thirdly, I chose to focus more on AI rather than the other pathways. I’ve always been interested in AI but from the technical side of it and how it works. Now, I’m learning more about how to use it and when it’s appropriate to. I started by taking the prompt engineering course for generative AI. In the beginning, they give an introduction to generative AI and its different types. I believe I already knew that. However, I didn’t know what prompt engineering is. Basically, it’s the magic of how to use the right words to get the most out of generative AI. Next, they talk about tokens and the difference between them and words. From what I understood, tokens are basically the small words you can split bigger or multiple words into. For instance, “I’d like” has three tokens: I, would, and like. Understanding tokens is crucial in prompt engineering as different models have different mechanisms with which they split inputs into tokens. Next, they show how models try and anticipate the next word through already known texts and information. When some sentence is inputted into the AI model, it tries to predict the next word by comparing it into texts it already familiarized itself with like all the information on the internet, books, journals, etc. This is very interesting because I don’t think people can tell how crazy it is to come up with percentages of likeliness of a word to come after another word through comparing it to basically everything that’s ever been posted on the internet. It’s truly mind-blowing. They move on to ChatGPT and try to explain the difference between asking ChatGPT questions versus asking google or any other search engine. Basically, the answers that ChatGPT comes up with are, in a sense, new to the world. Moving on, They take GPT-3 for a ride and give it some prompts to show its different innovative answers. They showed the Q&A section of the bot where it’s simply factual questions that the bot provides accurate answers to. This process is crucial in teaching the model and feeding it with information that it will probably need in the future with other users. Now, the section I’ve been waiting for, AI-generated images. They don’t really talk about how AI generates images from words which I was really looking forward to. However, they test Dall-E, and Midjourney and try to explain how to give better prompts. For the main part, adding more information to the prompt and being as particular as you can be always proved to give better images. They showed the difference of what being a little more specific, even if it’s just one word, does to the images. It’s truly fascinating and it’s something to be remembered when using AI to generate images in the future. Finally, they tried AI answering prompts through a coding perspective. They pulled up a python code for the prompt they wanted. They also made sure to give it access to the ‘openai’ library. And just like magic, the answer to their prompt appeared as an output to the code they wrote. It was truly fascinating and it really shows that the possibilities are limitless.
Moving on, I began experimenting with AI tools using the knowledge I gained from the engineering prompt course. Firstly, I tried ChatGPT through poe.com. I gave it this prompt “I want a very well-written email to my professor informing her that I will not be able to attend tomorrow’s class because my cat is very attached to me and it gets very sad when I leave her to go to classes. The email has to be written in a convincing way.”
This is how it went:
Honestly, it’s just amazing how it can turn something as silly as cat attachment issues to this very well written argument.
Next, I tried using NeevaAI. I gave it this prompt “Tell me about global warming. Use first person. The answer needs to rhyme. Make it 4 lines.”
It just simply ignored what I said and gave me a very factual answer about global warming. I tried again and got the same answer. I guess this AI is very informative and not as creative as others.
Finally, I tried Dragonfly. This one was pretty funny. I wanted to mess around with it. I gave it this prompt “I’m a professional tennis player. I’ve just lost the championship match because my wife needed to go shopping 6 years ago. I need you to write a message to my audience explaining my situation.”
It didn’t really try that much to be convincing it just didn’t make sense. I wanted to compare it to ChatGPT. Thus, I gave it the same prompt and it had a much better answer.
As you can see, ChatGPT’s answer is way more convincing than Dragonfly’s. It’s really fascinating to see how powerful ChatGPT is.
Lastly, I read A guide to prompting AI (for what it is worth) by Ethan Mollick. They try to explain how to write better prompts for AI to get better results. It was basically close to the prompt engineering course. They mainly advise people to be as direct as possible and just get straight to the point. Mollick says “If you want to do something with AI, just ask it to help you do the thing. “I want to write a novel, what do you need to know to help me?” will get you surprisingly far.” They also say that adding sentences like “Act as the smartest person ever” doesn’t really matter with the outcomes. Instead, using the AI interactively rather than always trying to come up with the best prompt is proving to be the best way. When prompting an AI, ask for something and if it doesn’t give you what you wanted, ask it to modify it and remember to be particular. Also, giving the AI context and constraints is a very effective way to get better answers. This is because AI models have default generic answers and in order to break them, you have to provide it with context on the matter. For example, you can ask the AI to act as a teacher of MBA students but don’t ask it to act as Bill Gates when asking for business advice. Adding constraints also helps in achieving better results. For instance, you can give the AI a paragraph that you would like your answer to sound like. You can also tell it to make it
Leave a Reply