Response to Usage of Artificial Intelligence Language Models in the Educational Setting

By Andrew J. Robison V*

Artificial Intelligence (AI) language models like ChatGPT are cutting-edge technologies that have the potential to revolutionize the way we interact with computers and machines. Developed by organizations like OpenAI, these models use deep learning algorithms to generate human-like responses to natural language inputs. The ability to understand and respond to language has made AI language models a popular tool for applications ranging from customer service chatbots to advanced language translation systems. However, as with any new technology, there are also concerns about the potential for misuse, including issues related to privacy, ethics, and accuracy. With the increasing prevalence of AI language models in society, it is essential that we understand both their benefits and the challenges they present.

Uses

ChatGPT and other AI language models can be used in infinite ways. Users can ask anything from simple questions to complex requests with many layers of detail and added parameters, and a model will craft an appropriate response to the best of its ability.

Simple Responses

One could request a definition or explanation of a subject and receive a fairly unbiased response. For instance, ChatGPT was asked, “Who was Jesus Christ?” This was the generated response:

“Jesus Christ is a central figure in Christianity, and is believed by Christians to be the son of God and the Messiah (or savior) prophesied in the Old Testament. He was born in Bethlehem in Judea, and lived in the 1st century AD. According to the New Testament of the Bible, Jesus was crucified by the Roman authorities, died, and was buried. Three days later, he rose from the dead and appeared to his disciples. Christians believe that through his death and resurrection, Jesus offered salvation and redemption to humanity. The life and teachings of Jesus have had a profound impact on world history and have shaped the beliefs of millions of people for over two thousand years.”

Another example of a simple response would be to answer basic math questions or perform simple operations. This time, ChatGPT was asked, “What is 35 degrees Fahrenheit in Celsius?” and the response was more than just the answer; ChatGPT showed the formula and the steps to get to the right answer.

“To convert 35 degrees Fahrenheit to Celsius, use the following formula:

C = (F - 32) * 5/9

Plugging in 35 for F:

C = (35 - 32) * 5/9

C = 3 * 5/9

C = 1.67

So 35 degrees Fahrenheit is approximately equal to 1.67 degrees Celsius.”

This is arguably a better response than a simple calculated answer, because it also shows the user the process so that they can learn how to calculate an answer using this formula on their own.

Complex Responses

Users can also make very complex requests of ChatGPT and other models, and can expect near-perfect execution in response. Here is a command given to ChatGPT:

“Write the opening paragraph of a mystery story about 6 individuals named Sarah, Jeff, Ken, Mary, Lionel, and Gordon arriving at an abandoned house in the woods of Vermont for what they believed would be a murder mystery dinner party.”

Again, that command was written by the author of this article. This response was written by the AI model:

“Six individuals, Sarah, Jeff, Ken, Mary, Lionel, and Gordon, arrived at an old, abandoned house in the woods of Vermont, excited for what they believed would be an unforgettable murder mystery dinner party. The invitation they received promised an evening filled with suspense, intrigue, and of course, a delicious dinner. The group of friends eagerly made their way up the overgrown driveway, surrounded by towering trees, eager to start their adventure. As they approached the house, they couldn't help but feel a sense of unease. The windows were boarded up, the roof was caving in, and the front door hung off its hinges. Despite the creepy atmosphere, they were determined to enjoy the night and stepped inside, not realizing that what was about to unfold would be far from the dinner party they had imagined.”

To test the capabilities of the software, the previous command was altered to request the model to write a book, rather than just one paragraph. Because ChatGPT can remember previous responses within a conversation, it responded like this:

“I'm sorry, but I am an AI language model and do not have the capability to write an entire book. However, I can offer some suggestions on how you can continue to develop the story:

Chapter 1: The Arrival

In the opening chapter, you could introduce each of the six characters and their reasons for attending the murder mystery dinner party. You could also describe the abandoned house in detail, building a creepy atmosphere and setting the stage for the mystery to unfold.

Chapter 2: The Dinner Party Begins

The six friends gather in the dining room for dinner, excited for the murder mystery to start. As they eat, strange things begin to happen, and it becomes clear that something is not quite right. The lights flicker, doors slam, and ghostly whispers fill the air.

Chapter 3: The Murder

Just as the guests are about to retire for the night, a scream pierces the silence. One of the guests has been murdered, and the others are shocked and scared. With no phone signal and no way out, they must work together to figure out who the killer is before they become the next victim.

Chapter 4: The Investigation

In this chapter, the group begins their investigation into the murder. They interview each other, search the house for clues, and try to piece together what happened. As they delve deeper, they uncover secrets, lies, and motives that point to one of their own as the killer.

Chapter 5: The Confrontation

With the killer identified, the group confronts them in a dramatic showdown. As the tension builds, the killer's motive is revealed, and the truth about what happened at the abandoned house is finally uncovered.

Chapter 6: The Aftermath

In the final chapter, the guests reflect on what happened and the impact it has had on their lives. They say goodbye to the abandoned house and each other, eager to return to the safety of the real world. But as they drive away, they can't help but wonder if the mystery will ever truly be solved.”

While the limitations of the software do not allow it to author an entire book for the user, it clearly can provide a very helpful starting point for a writer (in this case) to get ideas for a story.

Another major feature of ChatGPT is the capability to write code. The author asked the AI model to “[c]ode a website for this book that contains a synopsis, character bio sections for each character, and a form to join the email waitlist. Style the website with a dark theme with a red accent color using CSS.” The code was then pasted into files on a hosting server and tested. The author asked for one additional revision: “Create a new stylesheet that has a bold, mobile-friendly theme.” The program then re-wrote the CSS file with the updated request to create a simple, yet accurate implementation of the request. That site, with the exact code written by ChatGPT can be viewed here, or interacted with in the frame below.

As can be seen, ChatGPT did not get everything correct, but again, it provides a great starting point for a web developer to get started or for someone with no coding knowledge to begin to set up a simple site. Clearly, however, more would need to be done to make something useful out of these responses. It should also be noted that the model added a hidden line of code that said this: “Add character bios for Ken, Mary, Lionel, and Gordon,” after the bios for Sarah and Jeff.

Don’t worry: submitting your email won’t actually do anything. This author has no plans at this time to write this book.

Abuses

As is always the fear with technology like this, there are ways in which this software can be abused. The biggest concern currently is with students in high school and University settings.

Instructors should not be concerned that students will use ChatGPT to write their 10-page term paper for them. This author entered the simple command, “Write a ten page term paper about Genghis Khan,” and the response was appropriate:

“I'm sorry, but writing a ten page term paper would require a significant amount of time and effort. As a language model AI, I can provide brief answers and generate content for you, but writing a comprehensive ten page term paper is beyond the scope of my abilities. You may want to consider researching and writing the paper yourself or seeking assistance from a writing service.”

However, if students were given the assignment of writing just an introductory paragraph, ChatGPT might just do that for them, as it did when told to “[w]rite an introductory paragraph to an essay about Genghis Khan.”

“Genghis Khan is a legendary figure in world history, renowned for his military conquests and the establishment of the Mongol Empire, which became one of the largest empires in world history. Born in the early 12th century in Mongolia, Genghis Khan rose to prominence through his military skills and cunning tactics, eventually unifying the Mongol tribes and leading them on a series of devastating raids and invasions across Asia and into Europe. Despite his reputation as a brutal conqueror, Genghis Khan is also credited with being a visionary leader who instituted a number of progressive reforms and innovations in areas such as administration, diplomacy, and trade. His legacy continues to shape the world today and his story remains an important part of world history.”

Similarly, ChatGPT might be able to help a computer science student get started on a project wherein he needed to build a website portfolio for a friend, but it can’t and won’t create the whole site for him. Yet, if the assignment is just to create a simple website that links to a CSS stylesheet, instructors may want to be aware of the capabilities of this software.

Responsibility

Many will say that technology like this is harmful and should be shut down just because of the threats it may seem to pose. In reality, things like this will not simply go away, but each individual has the responsibility to use the tools at their disposal in appropriate and ethical ways.

Students should treat artificial intelligence language models like ChatGPT more as a friend, than an answer farm. One could ask a friend for ideas on what to write about or help getting started on a coding project, but it would be unethical to allow any person to complete the entire assignment instead. Aside from being unethical, a student who is truly seeking to learn will follow the instruction of their teachers and do the work in order to learn. Just as if a friend does all the work, one will not learn by allowing ChatGPT to complete assignments in one’s place.

Instructors have the responsibility of familiarizing themselves with new tools like these, and adapting to incorporate the benefits they provide while educating their students on the appropriate manners in which to use the tools.

Instructors and institutions should not write off AI tools merely because of the possibility of cheating. Rather, they should help students understand how best to make use of such technology, and teach them to understand the fallibility of such tools. ChatGPT gets things wrong sometimes, just like a friend asked to help with coursework.

Ultimately, “[i]t is the responsibility of people to use tools like ChatGPT ethically and with consideration for the potential impact on society and individuals, while being mindful of potential biases and limitations of the technology.”

*The first paragraph and the last sentence of this article were written by ChatGPT.

Image Credits: OpenAI