Hello everyone! I found this article about a computer science student at Princeton University, Edward Tian, who spent his holiday break in his hometown of Toronto developing an app, GPTZero, which can sniff out ChatGPT essays.
As we have discussed in class, the infamous generative AI known as ChatGPT has caused an uproar within many educational institutions, leaving some faculty (and students such as Edward Tian) worried about it jeopardizing the integrity of students’ or researchers’ work. On the other hand, others have embraced it, viewing the technological innovation not as a disruption but as a tool to advance learning (analogous to the introduction of Google or the calculator). The ChatGPT project’s initiative is to recreate the function of a Google search engine whereby the user inputs their question and the output is the most suitable answer filtered from sources off the internet. Where ChatGPT significantly differs is that it relies on an AI to generate a well-written answer to the question; adding a (somewhat) human element into the mix.
Fascinatingly, because algorithmic AI utilizes machine learning code, it will only get more efficient as it is being used. With more users, it makes for a better ChatGPT. In fact, last week ChatGPT passed the United States medical licensing exam (linked here) doing better than PubMedgpt (linked here). It also passed MBA courses, scoring a B to B minus at the University of Pennsylvania’s Wharton School (linked here). How long until it passes the bar exam? Should we be worried about it replacing our own jobs? What if AI replaces all jobs and no one no longer needs to work, and we can all run on Universal Basic Income? Alas, these are all big questions for a later date…
This article I found explains how Edward Tian was able to program his own AI app, GPTZero, that can trace the ChatGPT works written in students’ papers. Thus, in response to those faculty members who may be protesting for a blanket ban on ChatGPT, there is a resolution and it involves more AI (how fun!). Throughout its current testing the GPTZero has a less than 2% false positive rate. The article also explains how the AI works: testing two variables, perplexity and burstiness. Since its launch many teachers and admission officers have contracted for its use. If you wish to try it out for yourself, here is the link https://gptzero.me/. I threw in this post and the result was: “Your text is likely to be written entirely by a human”.
It may be worth noting that Edward Tian is not against ChatGPT. While he is concerned for its usage in academic plagiarism, he nonetheless encourages research and projects that foster the efficiencies of machine-learning artificial intelligence.
Overall, it is interesting to see how fellow coders can be the better-equipped response to public outcries caused by ChatGPT, hence displacing the need for reformative jurisprudence or legislative actions in regard to certain issues (such as questioning the originality in academic literary works which may be the subject of a copyright dispute, where an author allegedly used ChatGPT). This is not to say there is no room for the law, as many questions are still left unanswered (for instance, should an AI bot be entitled to its own copyright protection? etc) but simply to say that the world of AI is still in its early stages and sometimes more AI is born to evade (or diminish) the need for administrative bodies (like the CRTC’s role under the Bill C-11 proposal) or other quasi-policing bodies (like universities) to step in.