Student Develops Tool To Detect AI Written Essays
Students all around the world were probably psyched when the newest AI chatbot launched a few months ago because of its ability to write essays. But a Student develops tool to detect AI-written essays.
Although many chatbots have come before it, none have proven as successful as the revised ChatGPT released in November by Elon Musk's OpenAI.
The inventor of Tesla even made light of the notion that students will no longer need to do schoolwork in the future. It seems, however, that these hopes have been shattered since a student has purportedly developed an online tool that can determine whether or not essays are produced by the AI tool. Womp, womp.
Who Invented GPTZero?
English majors in high school who planned to have their assignments written by AI now have a new opponent. Tian, a computer science student with a journalism minor, spent some of his break developing GPTZero, which he claims can "quickly and effectively" determine if a person or ChatGPT produced an article.
COPYRIGHT_WI: Published on https://washingtonindependent.com/student-develops-tool-to-detect-ai-written-essays/ by Rhyley Carney on 2023-01-11T02:18:20.833Z
The rise of AI plagiarists is what inspired him to build the bot. Students have reportedly started exploiting the ground-breaking ChatGPT language model to submit AI-written homework as their own since its introduction at the end of November.
I spent New Years building GPTZero – an app that can quickly and efficiently detect whether an essay is ChatGPT or human written.
Tian added that since he put his bot online on January 2nd, many educators have been in touch to share the success they've had with their students using it.
After just one week, more than 30,000 users had already tested GPTZero. The software was overrun by users because of its popularity. Now that GPTZero's free hosting platform, Streamlit, is seeing more traffic, they have stepped in to help assist Tian by providing additional RAM and resources.
How GPTZero Works
GPTZero employs "perplexity" and "burstiness" as indications to evaluate whether an extract is authored by a bot. The perplexity metric is used to determine whether or not a piece of text is too difficult for a computer program like GPTZero to understand.
In contrast, the text is more likely to be created by AI if it is more familiar to the bot (due to its training on similar data), in which case it will be of low complexity.
To examine phrase variants in isolation, we may use the burstiness metric. For example, humans have a tendency to write more rapidly, alternating between shorter and longer, simpler and more complicated phrases. Most phrases produced by AI are quite standard.
Tian showed the app's analysis of a piece from The New Yorker and a post from ChatGPT on LinkedIn in a demonstration video. It was able to tell the difference between human and computer-written text.
Since his bot has been put to the test by users, Tian has conceded that it isn't 100% reliable. He promised that he was still toiling away at the model to make it more precise. But the app's design contributes to Tian's overarching goal of providing AI transparency by illuminating the differences between humans and machines.
Rise Of ChatGPT
Given ChatGPT's stratospheric development since its introduction, the student's claim that he has been approached by professors from across the globe is not surprising.
For instance, philosophy professor Darren Hick at Furman University was baffled to learn that one of his students had used artificial intelligence to avoid finishing her assigned assignment.
Today, I turned in the first plagiarist I’ve caught using AI software to write her work and I thought some people might be curious about the details.
The student used ChatGPT, an advanced chatbot that produces human-like responses to user-generated prompts. This technology is about three weeks old.
The Initiative To Combat AI Plagiarism
The senior in college is not alone in the effort to curb AI plagiarism and fabrication. ChatGPT's creator, OpenAI, has shown a dedication to combating AI plagiarism and other malicious uses.
Scott Aaronson, a researcher at OpenAI who focuses on AI safety, announced last month that the business has been developing a method to "watermark" GPT-generated text with an "unnoticeable hidden signal" to identify its source.
Hugging Face, an open-source AI community, has released a tool for determining if the text was produced by GPT-2, an older version of the AI model used to generate ChatGPT. A philosophy professor in South Carolina who was aware of the technology and utilized it to detect a student submitting an AI-written paper stated he learned about it by chance.
The New York City Department of Education said on Thursday that it is limiting access to ChatGPT on school networks and devices due to negative effects on student learning and content safety and accuracy issues. Tian has no objections to the employment of AI technologies such as ChatGPT.
[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.
It is not known, however, how effective GPTZero really is. According to at least one Twitter user, it missed some of their AI-written samples. The reception has been more divided elsewhere on the platform: Some adults have praised the initiative, while many others, especially young people, have called Tian a "narc."