OpenAI Text Classifier To Flag Content Written By AI

The popular ChatGPT reached a million users in just five days of its launch. You might have seen countless examples of how text-generating AI offers impeccable performance and it continues to be the case. Its maker OpenAI has released an AI text classifier that helps users detect whether the text submitted was written by a human or an AI tool such as ChatGPT itself.

Dubbed as OpenAI AI Text Classifier, the tool scans through the snippets to find whether the text was generated using AI or a human wrote the same. The tool has a success rate of 26% at the moment. OpenAI mentions that one should use the tool in tandem with others to gain resourceful insights. OpenAI is working on upgrading its performance and would deliver a robust classifier with successive iterations that will follow.

The official note further adds that one shouldn’t use AI Text Classifier as the only primary source for decision-making. Instead, they should use multiple methods to fine-tune their results based on the results. AI Text Classifier does have some limitations though. It requires at least 1,000 characters equivalent to 200+ words to carry on with the analysis.

Apart from that, it cannot scan for plagiarism which is a major issue with AI-based text generator tools. As TechCrunch puts it, text-generating AIs usually scour for information and regurgitate the text based on the training model. It makes sense to have a plagiarism checker but the classifier lacks it.

The classifier carries a success rate of 26% at the moment. It will tag texts that it finds to be completely written by humans or through AIs using the confidence level. With a less than 10% chance, the content is very unlikely to be AI-generated while those with 98% and above are likely to be AI-generated copies.

For now, the classifier has been trained on 34 text-generation tools hailing from five organizations including OpenAI owns ChatGPT, GPT-3, and DALL.E.

ChatGPT’s popularity skyrocketed just hours after it made its debut on November 30th. Stack Overflow has been riddled with dubious but convincing-sounding answers. Apart from that, many U.S. schools have banned ChatGPT on their networks while organizations around the world are taking their stand when it comes to AI-based text generators out there.