AI - Chatbot
New AI Classifier Improves Reliability in Detecting AI-Generated Text: Addressing Limitations and Mitigating Misinformation
Summary: OpenAI announces the launch of a new AI classifier for detecting AI-generated text. The classifier can distinguish between human-written text and AI-generated text from multiple providers and improve reliability in mitigating false claims and misinformation. The classifier has limitations and should be used in conjunction with other methods. OpenAI is seeking feedback and has developed a resource for educators on the use of AI systems such as ChatGPT.
As a leading provider of cutting-edge AI technology, OpenAI is proud to announce the launch of a new AI classifier aimed at detecting text generated by AI systems. The classifier is designed to distinguish between human-written text and AI-generated text from a variety of providers, providing valuable information for mitigating false claims and preventing the spread of misinformation.
Our new classifier represents a major step forward in the safe deployment of large language models. While it is impossible to detect all AI-generated text with 100% accuracy, our classifier has the potential to play a key role in addressing the challenges posed by the widespread use of AI-generated text.
In evaluations, the classifier was able to correctly identify 26% of AI-generated text as “likely AI-generated”, with an error rate of 9% for human-written text. The reliability of the classifier increases as the length of the input text grows and it performs significantly better on text generated by more recent AI systems.
It is important to note that the classifier should not be used as a standalone tool, but rather as a complement to other methods for determining the source of a piece of text. The classifier has limitations and can be unreliable on short texts, texts in languages other than English, and very predictable text. Additionally, AI-generated text can be edited to evade the classifier, making it a constantly evolving field.
OpenAI has made the classifier publicly available and is actively seeking feedback on its usefulness and limitations. Our team is committed to continuing to improve the detection of AI-generated text and will be sharing updated methods in the future.
To support the growing impact of AI-generated text on education, OpenAI has developed a preliminary resource for educators on the use of AI systems such as ChatGPT and the limitations and considerations associated with it. The company is actively engaging with educators in the US and welcomes feedback and resources from those directly impacted by these issues, including but not limited to teachers, administrators, parents, students, and education service providers.