The new AI software has begun to solve problems as some students consider its potential in completing assignments for them.
ChatGPT is a software that has taken college campuses by storm. This software helps students write code, come up with fake stories and even does their assignments for them with enough manipulation of the service. GPTZero is the service that can catch those who utilize this software. Edward Tian, a 22-year-old senior at Princeton University, built the app to detect if submitted work has been generated by AI. The app was created out of his own inspiration to detect human writing from AI-generated text.
The following paragraph is directly from ChatGPT, after asking the AI itself how it functions.
“ChatGPT is a transformer-based language model developed by OpenAI that uses deep learning to generate text. It’s trained on a large corpus of text data and learns to generate text based on patterns it observed in the training data. The model uses an attention mechanism to weigh the importance of different parts of the input when generating the output. Given an input sequence, the model generates a probability distribution over the vocabulary for each word in the output sequence and samples from that distribution to generate the final text.”
As described by the AI itself, it is an attention mechanism that translates to a cognitive-mimicking part of the AI. The AI tries to mimic human cognitive function and follows a probability sequence for how a sentence would normally be generated. But this function only mimics human speech patterns — it is not a 100% accurate copy of how a person communicates, and that is how Tian made his GPTZero software, by comparing the “perplexity” (complexity) and “burstiness” (variation) of text. If a sentence is high in complexity, then it is likely to be written by a human and not an AI, but if a sentence is low in complexity and is “familiar” to GPTZero, then that sentence will be flagged for a high probability of being written by an AI.
ChatGPT has become an important factor in homework, tests, quizzes and assignments in general, especially on TU’s campus. In anonymous interviews with students here at the university, multiple questions were asked to each student, including just how useful this software is, why the student has used it and what the student thinks the future of AI-generated homework looks like.
After conducting these interviews, there was a concentration of usage in majors that related to computers/high computer usage that was significantly higher than in majors that did not. This survey was not all-inclusive, so a statistically significant conclusion cannot be made regarding the numbers collected, but the data collected does show some sort of relation to computer usage and increased understanding and utilization of the ChatGPT service.
After interviewing roughly 15 students with varying majors, those interviewed who had a focus on computer science/engineering were discovered to have used ChatGPT more than any other major. Of those who answered yes to knowing what ChatGPT was, none of them outwardly admitted to using it to successfully cheat, although some admitted they attempted to use it for help with homework.
After these interviews, it is a fair assessment to say that students are going to use a service that allows them to complete homework easier or faster, whether professors want or allow them to do so or not. Another example of this is Chegg, which carries a large amount of information that is accessible with a minor payment each month, and allows users access to nearly infinite online answers with the opportunity to pose questions as well.
So why hasn’t Chegg come under the same fire that ChatGPT has? It is not exactly the same thing, but Chegg can be moderated by professors who write their own exams while ChatGPT comes up with answers for problems without necessary input from a professor. Due to this, ChatGPT can be used to cheat on exams and fill in the required work of a student or employee. This is why Tian wrote the GPTZero software. As quoted, “GPTZero is ‘not meant to be a tool to stop these technologies from being used… but with any new technologies, we need to be able to adopt it responsibly and we need to have safeguards,” which clarifies why professors want to get ahead of the potential risks of students using this service. By preventing cheating, attempted cheating, or any sort of misuse of this software, professors can limit the amount of academic dishonesty that occurs in and outside of the classroom, and GPTZero is another tool to help combat this issue.
