✍️ Column by Connor Wright, our Partnerships Manager.
Overview: ChatGPT has been the language model on everyone’s lips since its launch. Claims have been made that it will revolutionize the education sector, potentially spelling the end of students writing their papers. So, as a philosophy and Spanish tutor, I decided to see if ChatGPT would nullify my services.
Introduction
Since its late November debut, ChatGPT has been the large language model (LLM) on everyone’s lips. Not only has ChatGPT called into question the very existence of student essays in the future, but it has also got people thinking about how education is done in its entirety. On the one hand, the state of New York has banned the technology in public schools. At the same time, Italy has placed an outright ban on the LLM, citing privacy issues (which was reversed at the end of April 2023 after OpenAI provided a satisfactory response to the government agency). On the other hand, Ethan Mollick (an MBA professor) requires his students to use ChatGPT in their studies. Such claims are all made even though OpenAI admits this is just a model for conversation, albeit with immense qualities. It is not used to solve the world’s problems or write your essays for you. Nevertheless, the claims persist.
Given this backdrop, as a Spanish and philosophy tutor in my spare time, I wanted to see if ChatGPT would make me redundant. To do so, I quizzed the program on the two subjects I tutor, playing the role of the student trying to get help. First, I wanted to see how the program could help me with my Spanish.
ChatGPT as a Spanish tutor
To begin with, I was greatly surprised at the accuracy of the explanations the model gave me. Firing grammar questions from all angles, the model gave me detailed instructions about how to form tenses (see Figure 1 below on the conditional tense), offered examples about how to put the tense into use, and advised me why the sentence I typed was wrong.
Figure 1: ChatGPT shows me how to form the conditional tense
However, as a student, I wondered what would happen if I didn’t understand its explanation. When I said I didn’t quite comprehend its summary, it said, ‘No problem’ and offered me the same advice. This would not bode well with a student who is really struggling with the material. However, I cut ChatGPT some slack, given it had done well so far, and decided to move on.
In its explanation of the conditional tense, it used the quite sophisticated example of “si lloviera, cancelaría la fiesta,” which means “If it were to rain, I would cancel the party.” This phrase uses a ‘si clause,’ which brings in both the conditional tense and the imperfect subjunctive mood. Hence, I decided to probe into this example and asked the model to explain the tense “lloviera” because it does not look like a conditional verb. See Figure 2 below for its response:
Figure 2: ChatGPT’s response to my question about “lloviera” not being a conditional verb.
ChatGPT does not give itself enough credit: it was correct. However, my question has made the model rethink its answer, making it state a grammatically incorrect sentence with full confidence as the correct translation. As a student, I would take this sentence formulation as gospel, still not knowing the tense of “lloviera.”
To try and be helpful, I mentioned how the model was correct. Nevertheless, Figure 3 shows how it believed I was mistaken:
Figure 3: ChatGPT calls me out on my Spanish grammar
Unfortunately, ChatGPT is now convinced of the wrong answer. As seen before, a significant problem with the model being a tutor would arise here. A student needs to gain the advanced knowledge to question what ChatGPT would say, and the questions they may have are met with the same explanation. ChatGPT says with as much confidence what is true and what is not and offers you the same explanation.
As a result of the above, I think my job as a Spanish tutor is safe. However, what about philosophy?
ChatGPT as a philosophy tutor
I asked ChatGPT about the philosophy of Ubuntu, which is a big passion of mine and one I know well. As with the Spanish side of things, I was impressed at the information it could present about the philosophy. It mentioned how the philosophy values community and well-being, its focus on interconnectedness, and the popular translation of the Zulu “umuntu ngmuntu ngabantu” (I am because we are).
Like the clueless student that I was pretending to be, I wanted to ask some questions about the contents of the philosophy. I took the spiritual route and asked about the notion of ‘vital force,’ which appeared in some of the earliest writings about Ubuntu philosophy to represent a spiritual element that flows through reality, stemming from God. This helped convey the element of being in touch with ancestors and the community sharing the same spiritual space. However, ChatGPT did not appreciate my question, as Figure 4 shows:
Figure 4: ChatGPT does not like my question on ‘vital force’
The model was convinced that Ubuntu philosophy does not draw on “any specific metaphysical concept,” proceeding to give me the same explanation as before to justify its answer. I insisted that the philosophy did involve a spiritual/metaphysical element, which ChatGPT did not appreciate, as Figure 5 shows:
Figure 5: ChatGPT insists I am wrong (again!)
I was not going down without a fight and continued to argue that if the model didn’t want to bite at the use of ‘vital force,’ it was at least a spiritual philosophy. ChatGPT eventually surrendered and conceded that the philosophy has a “spiritual aspect,” as Figure 6 shows:
Figure 6: ChatGPT now includes spirituality in its repeated explanation of the philosophy
Once again, after insistent questioning, ChatGPT gets closer to the truth. Yet, this prodding wouldn’t result from a student. In my view, I think I can keep my job as a philosophy tutor too!
What the future might hold
I chatted with one of my Spanish students about using ChatGPT, highlighting how it can be a 24/7 service for basic grammar questions, given its relatively detailed explanations. However, the order of asking would be the following: their teacher, me, Google, a Spanish friend, and, finally, ChatGPT.
For what ChatGPT is intended to offer, it does a fantastic job and presents a really exciting future for LLMs. However, the danger comes when we use ChatGPT for purposes it was not designed for, taking whatever it says as the truth even when it is wrong. So, students, please continue to write your papers. However, if you need to get some inspiration or want to converse about your essay at 3 am, ChatGPT is at your disposal.