Soon after the release of the new ChatGPT bot, Professor Darren Hicks caught two students in the act of AI-giarism. The term “AI-giarism” refers to the practice of presenting content written by artificial intelligence (AI) as your own. Hicks claimed that he became concerned when the student submitted an essay that was on subject but contained some well-written false information. He ran it through an AI detector, which determined that the piece was 99% likely produced by AI. As AI technology develops, this practice is becoming more and more common. As a result of situations like this with students, ChatGPT has caused controversy in academia. Fear not, several organizations have constructed AI writing detectors to weed out AI-giarism and save the day for teachers.
GPT-2 Output Detector
The GPT-2 Output Detector is a tool created by OpenAI. Even though this software was created for the more outdated GPT-2 bot, it is still quite good at pointing out ChatGPT’s outputs. In fact, Darren Hicks used this software to confirm his suspicions. The detector will provide you with its conclusion a few seconds after you paste in the text you want to study. However, this program is more compatible with shorter literature that is less than 1,000 characters long. To ascertain whether a piece of writing was created by AI, this software delivers a proportion of a “real” to “fake” ratio. The only version of this tool that is accessible online is a demo, which is free. However, as ChatGPT develops further, this software will become much less useful.
Giant Language Model Test Room
The Harvard Natural Language Processing Group and the MIT-IBM Watson AI Lab collaborated to create the Giant Language Model Test Room (GLTR). This software focuses on a piece of writing’s complexity, assessing how likely a word is to be used. If the term is among the top 10 most likely options, it will be highlighted in green. Yellow highlights the top 100 words likely to be used, while red underlines the top 1,000. A word that is more unlikely to be used than that is highlighted violet. The majority of words in most written works will be highlighted in green, but because human writings are a lot less predictable than AI texts, they will also frequently contain a lot of words that are yellow, red, and even violet. However, the online version of GLTR was created in 2019, making it less reliable with the advancement of current ChatGPT outputs.
GPT Zero
Edward Tian, a senior at Princeton University, developed GPT Zero. The only model that is free online is an older beta model, although Tian has been working on a new model. GPT Zero measures the randomness in sentences and the randomness of the text as a whole. Human writing almost always contains more random content than text produced by AI because we tend to use words and sentence structures that are less predictable. As a result, an AI likely created the text if GPTZero detects low levels of randomness.
AI Writing Check
AI Writing Check is a completely free service created with educators in mind by two nonprofit organizations, Quill.org and Commonlit.org. They created AI Writing Check to make it possible for teachers to determine whether a piece of writing submitted by a student was written by ChatGPT. It allows teachers (or anybody else) to copy and paste text and in a matter of seconds learn whether the material was created using ChatGPT. However, you must divide the content into distinct sections to check responses longer than 400 words. Based on testing with 15k essays, they believe this tool is 80–90% correct most of the time. Nevertheless, they have advised teachers to use caution while utilizing this tool to look for academic dishonesty because of potential errors. They have described AI Writing Check as a band-aid solution that teachers can employ this academic year until more sophisticated AI detection technologies become publicly accessible.
AI Text Classifier
AI Text Classifier was created by none other than the creators of ChatGPT. On January 31st, OpenAI launched AI Text Classifier, which is user-friendly while still maintaining accuracy. All you have to do is enter at least 1,000 characters (about 150–250 words) of text into the box and press the submit button. The paper is then marked as highly unlikely, unlikely, unclear if it was created by AI, probably, or likely by AI. Users receive $18.00 of credit as a sort of free trial, and it lasts for around four months before expiring. After that, you pay for what you use. Prices depend on how many “tokens” are used. Tokens can be compared to bits of text, with 1,000 tokens equaling roughly 750 words. There are numerous models offered, Davinci being the most powerful, and Ada the fastest.
Application In the Classroom
Finding a detector for your classroom, or for anywhere else, might be a challenging choice with so many options available. Cost is a factor to be taken into consideration. Free software is still useful to some extent, but the best technology available will never be offered for free. Teachers must remember to exercise caution when using open-source software to uncover academic dishonesty because these programs are never entirely accurate. Regardless, these programs can be very helpful in catching an essay produced by an AI.