OpenAI’s AI Text Classifier has dramatically advanced the utilization of AI-generated content in several businesses. AI-generated text is currently on the rise and is immensely helpful in several operations where content is the priority. OpenAI wants companies to leverage the tool in an ethical and responsible manner to generate content.
OpenAI’s ChatGPT and GPT-3 models are generating huge amounts of content for various companies on diverse topics. That is the reason why the same company launches a tool in an attempt to distinguish between human and AI-generated content. It could be a valuable tool to prevent the abuse of AI text generators. However, OpenAI notes that it has a success rate of only 26%. The company feels that when used in tandem with other methods, it will help curb the increased use of AI-generated texts.
AI Text Classifier from OpenAI
Many organizations, especially school districts, are banning ChatGPT on their networks because they fear the potentially harmful effects of automated text generation on students. The impact of such a tool on student learning and accuracy may result in a dicey situation in the future. AI Text Classifier is trained on several examples available on the web to detect content generated by AI and separate it from human-generated text.
This new tool is a fine-tuned GPT model, as referred to by OpenAI, predicting how likely the content is to be generated by AI. Nevertheless, the company also hesitates while disclosing that the results may not be completely accurate as it has some limitations. Therefore, the results must not be taken as the sole evidence to decide whether it was AI-generated.
There are five labels of evaluation from the AI Text Classifier:
- Very unlikely (if the text seems to be less than 10% AI-generated)
2. Unlikely (if the text seems to be between 10% to 45% AI-generated)
3. Unclear if it is (if the text seems to be between 45% and 90% AI-generated)
4. Possibly (if the text seems to be between 90% and 98% AI-generated)
5. Likely(if the text seems to be over 98% AI-generated)
None of the labels show a tone of confidence in its final result.
Benefits of AI Text Classifier
The proliferation of AI-generated texts was harmful when it came to data sharing. Many organizations were unable to identify the difference between AI-generated material and human-written text, creating a hazy line of doubt and confusion on the reliability of the content. AI Text Classifier may not be 100% effective. However, it can curb the extensive use of AI-generated texts.
There has been an increase in other companies launching AI-generated text detector tools to help companies apply AI ethically and responsibly. Although it is not easy to solve the issue entirely, such detectors will still try their best to reduce the excessive use of AI tools that can be harmful to security researchers creating a loophole to help cyber criminals.
Takeaway
Most professionals were using AI to generate content to draft their proposals and other important documents. While some included their content and mixed the text up, many abused the tool by entirely relying on it. Many educators were unable to identify the works of their students and whether the assignment done by them was AI-generated or self-composed. AI Text Classifier is helping such professionals who are struggling to identify AI-generated content.
Although, OpenAI emphasizes that the tool hasn’t been tested extensively to detect content with a combination of AI and human writing. As a final note, AI Text Classifier can be a valuable tool for flagging AI-generated text, but it should not be relied on as a definitive measure.
For more such technology-related updates, visit www.onpassive.com.
