AI Detectors Are Failing To Detect Human Content Among AI-Generated Content

/ Artificial Intelligence, Stories /
Getting your Trinity Audio player ready...

Imagine this: you researched a topic, wrote a draft of it, edited it the next day, proofread the article with your friend, colleague, partner, or spouse, and then published it.

However, when you check the piece you have put your heart and soul into with an AI detector, it says it is “not human-generated”.

How would that make you feel?

Turnitin, a company that checks for plagiarism, has added a feature that can tell if someone is using AI writing tools like ChatGPT.

Teachers are using it to catch the people who are trying to cheat the system, and it has become one of the most sought-after AI detection websites.

Well, guess what? The results from the AI detection tool are incredibly misleading.

AI Detection Tools are Failing

A recent experiment tested 16 samples of essays written by five high school students and written by AI. Surprisingly, Turnitin wrongly tagged 50% of the test samples.

Since the launch of ChatGPT, Turnitin’s error rate has increased from 1% to 4%. 

Now, this is not only the case with Turnitin and other AI detectors such as Writer and ZeroGPT as well. None of these tools accurately detects if the content is genuinely written by a human or not.

Everyone is aware that AI is getting more and more updated with every release, and it is already getting difficult to find the authenticity of anything. 

Hence, people of all professions are using it to fact-check if any new content out there is machine-generated or not.

Professor fails the Students, Accusing them of Cheating

As a matter of fact, a professor at Texas A&M University failed the whole class after ChatGPT wrongly claimed that it wrote their papers.

Yeah, that’s what the agriculture professor, Jared Mumm, has accused the students of. He sent out an email to the students saying that he found out that the papers were written by ChatGPT.

“I put everyone’s last three assignments through two separate times and if they were both claimed by ChatGPT, you received a 0.” 

Mumm’s email

Unfortunately, some students who had already walked the stage at graduation were temporarily denied their diplomas due to this.

In fact, some of them have not even heard of ChatGPT. 

This sounds right as well. I know people who have heard about ChatGPT. It’s going to be six months since its launch.

Is US Constitution AI generated?

That’s what ZeroGPT displayed when someone used the prologue to the US Constitution.

According to the AI content detector, 92.15% of the United States Constitution was allegedly written by AI.

This tweet definitely sparked concerns about the effectiveness of AI detection tools.

Can AI-generated content indeed be detected?

My Thoughts

The rise of AI tools such as ChatGPT and Bard has undoubtedly led to a growing demand for these AI detection tools.

However, detecting AI might be as difficult as finding a needle in a haystack!

Now, the United States is deeply thinking about regulating AI. Italy, on the other hand, has blocked access to ChatGPT, while others in Europe might follow. 

Whatever the case, I just hope we get more powerful AI detection tools than what we might already have. Otherwise, there would be no stopping the increase in crimes and whatnot.

Related Posts