GPTZero is a tool that aims to determine whether a piece of text was written by a human or an AI. But does it work?
The tool works on the assumption that human-written language exhibits properties of burstiness: non-common items appear in random clusters that will certainly appear over time. Recent research has extended this property to natural language processing. “Some human written sentences can have low perplexities, but there is bound to be spikes in perplexity as the human continues writing. Contrastingly, perplexity is uniformly distributed and constantly low for machine generated texts.”
I tested GPTZero with some AI-generated text, and it came to the conclusion that my AI-generated text was written by a human!
While GPTZero was unable to accurately distinguish between the human- and AI-generated text in my test, this does not necessarily mean that the tool is a failure. Instead, it highlights the complexity of the task and the importance of continuing to research and develop new approaches to natural language processing and