Wednesday, February 21, 2024

Can’t rely on ZeroGPT to detect AI generated content as it is not accurate

Share

In this evolving world of AI, people have turned out to be very dependent on Artificial intelligence. Everyone has adopted a lifestyle in which AI is playing a very important role. There are a lot of things that AI can do well for society, it is doing indeed but there are a lot of things that it is not doing for the betterment of the generation and the upcoming ones.

In many surveys, it is noted that children are using AI to complete their homework and assignments which is not a great thing or something that needs appreciation. Students can surely take references but, it is not okay to use AI to complete the entire assignment. It is not only killing the creativity and interest to work hard in children but also it is making this generation super lazy! The assignments that use to take about 4 to 5 hours to get completed are now being done only in hardly 10 minutes.

Not only it is about assignments, but it is now highly being used in the commercial sector to finish the work fast and more efficiently. But, there are also drawbacks to it. The work is being done very fast but the thing that is left behind is the skills of the person. That person will soon be addicted to work only with the help of AI, and in the absence of AI, won’t deliver high-quality content. This will surely degrade his/her skills.

OpenAI, the creator of ChatGPT, the universally used and loved chatbot, has also created software which claimed to detect AI-generated content. Initially, teachers and other people were excited to use it to detect the homework or assignments which students completed using AI, but found out that it is not at all accurate.

OpenAI claims it is 98% accurate but, it is not at all accurate. It cannot detect if the text is generated by AI or not. To figure out this and to see if this rumour, team Sociobits conducted a basic testing of ZeroGPT.

In this test, we asked ChatGPT to write us a short article on Gaming. It surely gave us a good article along with the title and also subtitles. It was well-structured and was very properly drafted.

We copied one paragraph and pasted it on ZeroGPT to find out if it is accurate or not.

And to a surprise, it was NOT ACCURATE!

Here it showed us that the content is 78.07% AI generated, but as you all know it was entirely generated by ChatGPT! There was not a single change in the text done by us and still, it showed us that it was not entirely generated by AI.

Additionally, it highlighted the text which was generated by AI and did not highlight a few sentences. But as far as you all know, everything was generated by AI.

They say “After analyzing more than 10M articles and text, some generated by AI and others written by humans, we developed ZeroGPT’s algorithm with an accuracy rate of text detection up to 98%. Our AI text detector tool uses DeepAnalyse™ Technology to identify the origin of your text.
Our experiments are still ongoing, and our aim is to analyze more articles and text, and to converge to an error rate lower than 2%.” But, now we know how true it is!

Despite the significant amount of attention it has received, our testing reveals that GPTZero does exhibit some accuracy in distinguishing text generated by ChatGPT from random guesses. However, it is important to note that it frequently makes incorrect judgments. In the context of allegations of educational misconduct, where plagiarism can result in severe consequences such as failing grades or expulsion from academic institutions, this level of accuracy is insufficient and raises concerns.

Read more

Recommended For You