OpenAI, the AI research firm behind ChatGPT, has released a new tool – the AI Text Classifier – to help distinguish between AI-generated and human-generated text. This tool is intended to help mitigate false claims that humans wrote content generated by Artificial Intelligence (AI). According to OpenAI's announcement, it can limit the ability to run automated misinformation campaigns, use AI tools for academic fraud, and impersonate humans with chatbots. Although it cannot detect AI written text with 100% accuracy, when tested on a set of English texts, the tool could correctly say if the text was written by AI 26% of the time. However, it also mistakenly thought that human written text was written by an AI 9% of the time. OpenAI admits that its tool works better with longer texts; thus, it requires a minimum of 1,000 characters for running a test.
OpenAI's new AI Text Classifier is a powerful tool, but it also has some limitations. For instance, it can mislabel both AI-generated and human-written text, and minor edits can help AI-generated text evade the classifier. In addition, it was primarily trained on English content written by adults and may not be as reliable with text written by children or in languages other than English.
To test the OpenAI AI Text Classifier out for yourself, log into platform.openai.com/ai-text-classifier, paste the text you want to evaluate, and hit submit. The tool will provide an evaluation of how likely it is that the text was generated by an AI ranging from very unlikely to likely or possibly unclear—all within seconds! To illustrate this in action, I tested ChatGPT's ability to generate an essay about SEO and submitted it verbatim to the classifer; it rated the essay as possibly generated by AI. OpenAI developed a tool called the AI Text Classifier which is designed to distinguish between human-written and AI-generated text. This result from January 2023 (https://platform.openai.com/ai-text-classifier){rel="nofollow"} illustrates the tool's limitations, as it could not say with a high degree of certainty that the ChatGPT generated text was written by AI. By applying minor edits suggested by Grammarly, however, I reduced the rating from possibly to unclear. OpenAI is correct in stating that it’s easy to evade the classifier, but it is not meant to be the sole evidence when deciding whether a document was generated with AI (https://platform.openai.com/faq#what_is_the_intended_use_of_the_tool){rel="nofollow"}. OpenAI adds that this tool has yet to be thoroughly tested to detect content containing a combination of AI and human-written text, as such they suggest relying on other sources when determining if a document was written by AI or humans.
The AI Text Classifier is a powerful tool for identifying potential AI-generated text. However, it is important to note that the classifier should not be used as a definitive measure for making a verdict. Instead, it can provide helpful insights into how much of the text might have been produced by an artificial intelligence program.