Partner with Student Hires for the 22-23 School Year!

University Student Develops GPTZero, a Tool to Identify AI-Generated Written Pieces

Estimated Reading Time: minutes

Last Updated on January 12, 2023 by Deandre Barrett
GPTZero

Written By Deandre Barrett  |  News  |  0 Comments | January 12, 2023

Students hoping to use ChatGPT for their papers with little effort might face some problems these next few months because of GPTZero. For example, Edward Tian, a Princeton computer science student, created GPTZero. This app can “quickly and efficiently” determine whether or not a person or ChatGPT produced an essay by exploring the “perplexity” and “burstiness” of the work.

He spent part of his winter vacation developing the program and decided on Jan 2 to unveil it to the public for testing and hopefully help educators worldwide.

The goal of the application is to combat what he perceives as an increase in “AI plagiarism.” Students have been using ChatGPT’s revolutionary language model to pass off AI-written work as their own since its release late last year.

Tian said that after he released GPTZero online on Jan 2, many instructors reached out to him, telling him about the excellent outcomes they’ve seen from testing it. Then, within a week of its launch, more than 30K individuals had downloaded GPTZero.

The software crashed as a result of how popular it was. Streamlit, the platform where GPTZero is hosted for free, has stepped in to help Tian by increasing his RAM and resources so that he may better handle the web traffic better.

How Can GPTZero Detect AI-Generated Writing?

 GPTZero compares human writing and AI with two distinct indicators: “perplexity” and “burstiness.” The complexity of language is measured by perplexity; if GPTZero is confused by the material, it’s more likely to be written by a person.

However, if the bot has been trained on such data, its text will have low complexity and more likely to come from an AI program.

Edward Tian showcasing different aspects of ChatGPTZero

GPTZero can learn more and be trained similarly to ChatGPT. Burstiness compares the changes in sentences.

Humans prefer to write with more burstiness, such as with longer or more complex phrases combined with shorter ones. AI-written language is generally flatter.

However, the technology must continually update and match different programs’ complexity. Students will work to find ways to deceive this programming. AI developers aim to make their programs smarter but also more human.

As technology improves, it might be harder to detect certain things without more intelligent detectors. However, programs similar to Tian pave the way for more developers to join programs and contribute to the movement.

What Will Happen to AI in Education in the Future?

While educators have no plans to pull the plug on any open-source AI projects, there needs to be some regulation on how students and teachers can use them. For example, the developer of ChatGPT, OpenAI, has emphasized a commitment to preventing AI plagiarism and other bad uses.

Scott Aaronson, an AI researcher at OpenAI, is currently working on “watermarking” GPT-generated text with an “unnoticeable secret signal” to identify its origin.

Many schools are still determining whether or not they want to ban the application temporarily through the school servers or restrict specific access to the program for teachers to use.

While school districts in New York plan to block the program entirely, some schools in Florida plan to monitor the situation and see how everything plans out.

For example, the open-source AI community Hugging Face has produced a tool that checks whether the text was generated by GPT-2, an earlier version of ChatGPT’s AI model. According to a South Carolina philosophy professor who was aware of the software, he used it to catch one of his students submitting work created by artificial intelligence (AI).

Students are finding ways to combat this programming here with this Reddit user. It will be an arms race between developers, teachers, and students regarding these new programming tools. Despite these challenges, AI is going nowhere soon. Schools must keep up with the recent changes to ensure their students are using the system effectively.

Student Hires and AI

Student Hires is dedicated to helping disadvantaged youth in our community increase their career readiness and expand their learning with our K-12 expanded learning program led by college students. 

We will monitor the upcoming discussions regarding AI platforms similar to ChatGPT and see how the benefits and negatives will affect our platforms. We always put our students first in our after-school programs and want to ensure their growth isn’t affected. If you have any questions, please get in touch with us for more information.

Conclusion

In conclusion, ChatGPT is an AI website students use to make essays. Real people write essays, but artificial intelligence assists in the writing process. Many teachers are concerned about this website because it is used to cheat on essays and math assignments. GPTZero aims to help assist with stopping students from doing that. So what do you think about ChatGPT? Do you believe it should be banned in schools or restricted somehow?


Last Updated on January 12, 2023 by Deandre Barrett

Deandre Barrett

Deandre Barrett is a computer programming major at Lehigh Carbon Community College. He currently juggles a life balance between doing course work and marketing apprenticeships with Acadium. After graduating from the Acadium 3rd cohort in 2020 and finishing creating gaming reviews for Blasting News in 2017. He is now creating content for Student Hires and looks to use his experience to take the company to the next level. Student Hires has been focused on collaborating with K-12 schools & universities, as well as community employers, to create valuable job opportunities for local college & university students.

Leave a Comment

{"email":"Email address invalid","url":"Website address invalid","required":"Required field missing"}
>