- A tool to identify text generated by AI has been made available by the nonprofit research organization known as OpenAI.
- It is acknowledged by the team that it is impossible to reliably determine whether a text is written by an AI or a human every time but it can still have its uses.
- The tool can be used to detect academic fraud, spot automated disinformation campaigns, make sure a human being is behind the customer service text, and more.
OpenAI, an AI research company, has released a tool to detect AI-written text. The tool is designed to help users identify whether the content was written by a human being or an AI. The software analyzes the language and style of the text in order to make its determination. The tool can be particularly useful for teachers, administrators, parents, students, and those who provide educational services.
Not fully reliable, yet
The tool was refined using a dataset comparing human-written texts and artificial intelligence ones on the same topic. To reduce the false positive rates, the OpenAI team changed the confidence threshold which means that content is only identified as AI-written if the tool is very confident.
OpenAI said in the blog post announcing the news:
« Our classifier is not fully reliable. In our evaluations on a “challenge set” of English texts, our classifier correctly identifies 26% of AI-written text (true positives) as “likely AI-written,” while incorrectly labeling human-written text as AI-written 9% of the time (false positives). Our classifier’s reliability typically improves as the length of the input text increases. Compared to ourpreviously released classifier, this new classifier is significantly more reliable on text from more recent AI systems. »
The OpenAI team acknowledges that it is impossible to reliably detect whether a text was written by an AI or a human being every time but they believe that the tool could be useful in specific situations, such as when people are running automated misinformation campaigns, using AI tools for academic fraud, positioning AI chatbots as humans, or just making sure the customer service behind a company is run by actual humans or not.
While the tool is publicly available, you need to enter your phone number to sign up and to be able to test it. We had an interview with ChatGPT, OpenAI’s AI chat tool, and used that text to see if it could identify its own writing. Here is the result:
You can test the AI classifier tool by clicking here.
- Also, you may be interested in: What is ChatGPT? Everything you need to know