Listen to Podcast:
The people who made ChatGPT, OpenAI, have heard criticisms about its AI chatbot. Namely, those from teachers who were worried that students were using ChatGPT to get their homework done or papers written for them.
On Tuesday, OpenAI released a new AI classifier tool (Opens in a new window) that is meant to directly address these worries. OpenAI made a free web tool to help people figure out whether a block of text was written by a person or a computer.
It’s easy to use the tool. Just go to the AI classifier, login, and copy and paste the block of text you want to check. The classifier will then put the text on a scale from likely to very unlikely to show how likely it is that it was made by AI.
#FPTech2: Schools and universities have been warning that students would use #AIgenerators like #ChatGPT to cheat in exams and assignments. To curb that, #OpenAI has launched a new tool called #AITextClassifier.https://t.co/ZDwz3JHeMU
— Tech2 (@tech2eets) February 1, 2023
Even though this seems like a great way to address some worries about AI writing text, OpenAI makes it clear that there are some caveats. For instance, OpenAI says that “the classifier is not completely reliable.”
“In our tests on a “challenge set” of learning English texts, our classifier correctly identified 26% of AI-written text as “likely AI-written” (true positives), while it mislabeled 9% of the human-written text as AI-written (false positives),” the company said in a blog post (Opens in a new window).
In its announcement, OpenAI talked about some of the other problems with the classifier. For instance, the classifier does a much worse job with text that has less than 1,000 words. It can sometimes mistake text written by a person for text made by a computer. And it works best with text written in English.
OpenAI isn’t the first company to make a tool that can help tell when the text was made by a computer. In fact, it’s not the first time the company has made a tool to help people sort things. OpenAI plans that this new tool is “significantly more reliable” than the one it replaced. It’s also interesting that one of the leaders in AI is constantly working on this kind of tool.
Still, there is a long way to go before AI-detection tools can be relied on by teachers and educators to find cheating in the classroom.
