The Potential End of Education

By Sarah Howard

Artificial intelligence (AI) leading to the end of the world is a prevalent theme in pop culture. In the movies, it seems like such a far-fetched idea, but in reality, we may be closer to an AI reckoning than we think. Today’s AI isn’t going to ruin us by crashing a meteor into the earth or killing everyone like they do in the movies. Instead, it is slowly invading our education system and harming our youth’s skills. If left unchecked, AI can and will harm the abilities of future generations.

The AI that has received the most attention in recent months is called ChatGPT. ChatGPT is an AI model that is able to answer simple questions. For example, if you ask it, “Who was the first president of the United States?” the AI will respond by saying, “George Washington was the first President of the United States. He served as the country's first President from 1789 to 1797.”

Every AI, including ChatGPT, is developed by a process called machine learning (ML) which can be divided into many different types of ML. The initial type of learning they used is called Reinforcement Learning from Human Feedback (RLHF). In this scenario, the developers wrote their own back-and-forth conversations between people and had the AI study them. This “taught” the AI how to copy the way people talk to eachother.

What gives ChatGPT a seemingly infinite amount of knowledge is called its training data. Every AI has a set of training data which is essentially a big list of examples for the AI of how it should behave or what it should say. In the case of ChatGPT, the training data is the entirety of the internet. Not only has it learned facts and behaviors from its developers, but it’s also learned facts and writing styles from all over the internet. It could even use this blog post to train itself how to write like me.

This leads to one of the primary concerns of the developers of ChatGPT. They worry that the AI may be inaccurate when answering questions because it may gather facts from websites with inaccurate information. When you open the chat with the AI, there is a warning that says it may generate inaccurate information.

For example, if you ask ChatGPT, “When did Queen Elizabeth II die?” the algorithm says, “Queen Elizabeth II is currently still alive as of Jan 2023.” Unfortunately, this “fact” isn’t true because Queen Elizabeth II passed away at the end of 2022, ending her historic 70-year reign of England.

Due to the sheer size of the training dataset, it would be nearly impossible for the developers to ensure every “fact” that the AI generates is actually a fact. It would require fact-checking the entire internet which is not feasible. This raises concerns for educators everywhere. This AI is advertised all over social media as a way to make doing homework easier and faster, but people fail to mention the dangers of false information.

However, the aspect of ChatGPT that is most concerning for educators is that it can write eerily human-sounding essays and articles. In fact, as an experiment before writing this post, I asked ChatGPT to write an article on the dangers of AI in education and it wrote a section on the concern of inaccurate facts. It said…

“One major concern is the possibility of generating misinformation or fake news. ChatGPT, like other language generation models, is trained on a large dataset of text, which may include false or misleading information. As a result, the model may generate responses that are not entirely accurate or true. It's important to fact-check any information generated by the model before sharing it with others.”

This shows the true danger of AI. It’s able to mimic human writing so accurately that it could fool many people. If I had used that paragraph or any paragraph generated by ChatGPT, it is likely that people wouldn’t be able to distinguish between my writing and the AI’s.

Since this AI has been so widely shared among students, teachers are concerned that students will be tempted to just type their essay prompts into the AI and submit the result as their own essay. If students become over-reliant on AI generating their essays, it would be detrimental to education everywhere because students wouldn’t practice the writing skills they learn in classes.

This is a concern that was addressed by a college senior named Edward Tian who is a senior at Princeton University majoring in computer science. When he first saw ChatGPT’s abilities, he was obviously amazed by how impressive it is.

Tian also thought of the consequences of this AI being publicly available to fellow students. He set out to create an app to detect when something has been written by an AI. The app, called GPTZero, uses the machine learning of ChatGPT against itself. The AI in the app can basically say, “Yes, this is something I could create,” or, “No, this would require more knowledge than I have to create.”

ChatGPT remains a threat to education because this app is relatively new, it’s still not perfect, and it’s not widely used among educators. We also have to keep in mind that AI is constantly evolving because developers will continue to feed ChatGPT with more data. If the tools to detect AI-generated text don’t evolve at the same rate, they won’t be able to tell when AI has been used to write.

Since, for now, people are still smarter than AI, it is up to us to prevent AI-generated essays from becoming the norm.

Parents, if you notice that your student is spending less time on their homework than normal, try to figure out if their speediness is a new-found skill or the use of AI. Have discussions with them about the value of practicing what they’re learning in class.

Teachers, to keep your students from using AI in their essays, first look at their past writing. If you notice that they have drastically improved their skills in a short period of time, it may be time to ask them if they know about the power of AI.

Another way to test whether or not a student is using AI to cheat is to have in-class writing assignments. If you compare their in-class work to their homework and the homework seems to be much better than what they produced under your supervision, that can also be a sign that they are using AI.

One of the benefits of Writing Symphony is that our teachers work on students’ writing in class and are easily able to compare their in-class work to their homework. This reduces the likelihood that our students will use this workaround because we require them to present and understand what they’ve written and we require them to write in class.

Lastly, for any students reading this, please do not use AI to write your essays. It may be tempting to just put your essay prompt into ChatGPT, but without brainstorming and writing yourself, you will never learn what your teachers are teaching you. You won’t be able to receive accurate feedback on your skills and you won’t have practice writing. Two of the primary ways we learn are by receiving feedback on what we’ve written, and by writing facts and thoughts to explain them through our essays.

AI is a powerful tool that isn’t just a Hollywood plot device anymore. If it goes unchecked, the power of machine learning will degrade human learning around the world. It is up to us to limit its use to maintain our knowledge and our skills before it’s too late.

Works Cited:

Goldstein, Josh A, et al. “Forecasting Potential Misuses of Language Models for Disinformation Campaigns-and How to Reduce Risk.” OpenAI, OpenAI, 11 Jan. 2023, https://openai.com/blog/forecasting-misuse/.

Marr, Bernard. “ChatGPT: Everything You Really Need to Know (in Simple Terms).” Forbes, Forbes Magazine, 21 Dec. 2022, https://www.forbes.com/sites/bernardmarr/2022/12/21/chatgpt-everything-you-really-need-to-know-in-simple-terms/?sh=17f70dc4cbca.

OpenAI, OpenAI. “CHATGPT: Optimizing Language Models for Dialogue.” OpenAI, OpenAI, 30 Nov. 2022, https://openai.com/blog/chatgpt/.

Rosalsky, Greg, and Emma Peaslee. “This 22-Year-Old Is Trying to Save Us from Chatgpt before It Changes Writing Forever.” NPR, NPR, 17 Jan. 2023, https://www.npr.org/sections/money/2023/01/17/1149206188/this-22-year-old-is-trying-to-save-us-from-chatgpt-before-it-changes-writing-for