OpenAI warns teachers that ChatGPT cannot detect AI-generated content, but offers tricks for identifying copied answers

Rate this post

OpenAI has clearly informed educators that no tool including ChatGPT is capable of detecting whether content is AI-generated. In the teacher-specific FAQ section, OpenAI stated that it has built tools to detect whether content is AI-generated or human-generated, but that they are not completely reliable. Teachers are made aware of this because many students tend to write assignments and homework using these tools.

According to the company’s FAQ section, “While some (including OpenAI) have released tools intended to detect AI-generated content, none of these have been proven to reliably distinguish between AI-generated and human-generated content.”

Speaking about its own AI chatbot, OpenAI revealed that it has “no knowledge” of what AI-generated content might be. Even if you ask the chatbot about it using different prompts, these responses will be random and are explained as having “no basis in reality”.

It further explained that when they tried to train their AI tools to detect AI-generated content, it sometimes labeled content like Shakespeare and the Declaration of Independence as AI-generated. Additionally, OpenAI stated, “Even if these tools can accurately identify AI-generated content (which they cannot yet), students can make small edits to avoid detection.”

How to find copied assignments?

To combat this problem, OpenAI has released some guidelines that educators can follow. To find out if students have copied their assignments from ChatGPT, OpenAI suggests that teachers can ask unique questions related to the content and see the relevance of the answer received and how well they understood the assignment. They still need to double-check and consider potential biases in that information.

It added, “Sharing interactions with the model ensures that students are held accountable for the use of AI in their work. Teachers can check that students are engaging responsibly and meaningfully with the tool, rather than just copying answers.”

Interestingly, according to OpenAI guidelines, users must be at least 13 years old to use ChatGPT. Users 13-18 years of age must have parental or guardian permission to use ChatGPT.

Also read:

Twitter turnaround: From audio-video calls to X Premium, Elon Musk’s biggest changes to X so far

Google Search Gets Generative AI in India: How to Enable and Use it in English and Hindi

Apple Wanderlust event: What to expect from iPhone 15, iPhone 15 Pro to Apple Watch Series 9

Leave a Comment