Lucas Horsman

AI: A fad or here to stay?

07 May 2024

I. Introduction

Artificial Intelligence has undergone a massive growth period in the last few years following the release of several Large Language Models for public usage. These include OpenAI’s ChatGPT, Google’s Bard, and models like Devin or LLaMA. While AI has been around for a while, these models specifically, as well as other “text generators’’ hold great significance in the field of education, specifically educational fields of computer science. Not only are they good at generating lots of code quickly, they are also prone to massive errors and blind spots if left unchecked. Not to mention the potential problems with plagiarism that come with a machine that gives 75% accurate answers to just about any question. Personally, I feel that AI is a crutch, and, while I do use it rarely, when I use it, I know in the long run I am damaging the learning process.

II. Personal Experience with AI:

I have used AI in class this semester in the following areas:

  1. Experience WODs e.g. E18 I did not use AI for this, as if I had issues, I could simply either try and figure it out (which usually resulted in me actually learning the material) or viewing the screencasts if absolutely necessary.
  2. In-class Practice WODs I challenged myself to not use AI as I wanted to make sure I actually learned the material.
  3. In-class WODs I used AI during these sparingly. I found the majority of them to be really easy to just read the problem and pretty quickly understand what I needed to do. Since I learned the material throughout the week, I was well prepared for the WODs. The few times I did, I would put in the problem(i.e mass copy/paste), ask it to begin generating the answer in the background, then simultaneously read the problem on my own. Then, if I had the same idea for the solution as was generated, I would just tweak the generated code to ensure the problem was solved correctly. If it didn’t have the same general idea as me, I would simply move on, as I do not enjoy asking it questions / providing instructions to push it in the right direction
  4. Essays I only used AI to verify grammar / spelling. Even still, I’d usually just use something like MS Word or Google Docs for this.
  5. Final project I wanted to implement some meteor methods to do things with my database that was not covered in class. ChatGPT helped me figure out what I needed to do and what documentation to read to learn this material.
  6. Learning a concept / tutorial Like above, I did find ChatGPT pretty useful to do something like this. If I came across a line of code I didn’t understand, AI could provide a myriad of situations that line would appear in, giving me additional context and understanding.
  7. Answering a question in class or in Discord I never found this necessary. If someone was asking me a question, they were likely looking for my specific input, and as such, using AI to answer the question would diminish my perspective on their problem.
  8. Asking or answering a smart-question Again, not necessary.
  9. Coding example e.g. “give an example of using Underscore .pluck” See 6.
  10. Explaining code See 6.
  11. Writing code I did not use AI to write code outside of the above scenarios. Often I already have an idea of what I want to do and how to implement it, so I did not need it for writing code. There were some cases where I would use AI to generate boilerplate HTML or Bootstrap, so I could just get to work on implementing the specificities of the problem.
  12. Documenting code I would ask ChatGPT to analyze and comment code in certain scenarios, like if I did not want to manually write comments. This still required making sure what it was saying about my code was correct.
  13. Quality assurance I would occasionally ask, out of curiosity, if there was anything I could be doing better when using specific functions (like react/mongodb components). Usually it would say I would run into security concerns or something of the like, but the suggestions were outside the scope of the ICS 314 course and my current knowledge.
  14. Other uses in ICS 314 not listed above There are no other uses not covered above.

III. Impact on Learning and Understanding:

Overall, AI has had a pretty minor impact on my learning experience. Everything I used it for could just as easily be solved by going through a few Google searches.

IV. Practical Applications:

Honestly, I have not done any research concerning practical usages of AI, and as such I am not knowledgeable on this topic.

V. Challenges and Opportunities:

I didn’t really run into problems. I either found it to be useful or not as described in the cases above. In education, I don’t see how AI could be used for positive growth when compared to other, tried and true methods like asking questions or reading material.

VI. Comparative Analysis:

In my experience I feel like something we should have learned over COVID is that learning is done better in person. Even something like a flipped classroom is very conducive to my learning environment. If we were to run into using AI as a classroom tool, I could see myself becoming frustrated with it very easily. I could also see myself learning to rely on it rather than actually struggling with the material and learning it.

VII. Future Considerations:

I think as AI develops, we will see more and more of it, until a certain point. Eventually, the bubble is going to burst, so to speak. Right now, I feel that outside of very sparse use cases, LLMs like ChatGPT are hardly impressive, and overall not very useful, both as a consumer and as a software engineer. Something that would need to change drastically is its propensity to generate incorrect information. Generally, if asked something that could be found on Wikipedia (i.e “what is the penguin’s predators?”) I still feel the need to go fact check it. This is something OpenAI is aware of at least with the warning message under each text box. I’m not even sure if this is something they are concerned about as they develop ChatGPT (at least in its current state).

VIII. Conclusion:

Overall, I think AI has some use cases, both in the consumer and in the software engineering space. However, in my experience, I have generally found it to be slower and less conducive to my learning experience when compared to asking questions to a human or simply using google searches.