A post on TechSpot notes that a tool has been created to counter the use of ChatGPT in plagiarism. The tool, called ‘GPTZero’, was created by student Edward Tian (22) at Princeton University and uses two key signals to find out if the AI-tool ChatGPT was used.
One signal is called ‘perplexity’ and the other is called ‘burstiness’. The first one, perplexity, will check the complexity of the text – in this case how much GPTZero is perplexed by the text. High perplexity? Very probably human-made. Low perplexity? That’s got AI written all over it – GPTZero will then have found things like text patterns to label it AI-created.
Burstiness as a signal looks at sentences and how much variation is in there: variation is a key point when humans create sentences, varying between longer sentences and shorter ones. Edward Tian notes that GPTZero is far from perfect (the tool is still in beta), but should be able to perform a quick check when a text is suspected as being created with ChatGPT.
ChatGPT
Usage of ChatGPT, an AI chat-platform by OpenAI has been on the rise recently with it often being ‘at capacity’ showing a lot of usage worldwide. Microsoft has been noted that it intends to invest $10 billion in the company, to be able to implement it in its portfolio of office software for example.
