The AI application ChatGPT has gained quite a bit of attention recently. It has an incredibly wide range of applications, from coding to answering queries and translating — all in an easy-to-use chatbot format. One of the most significant applications it has is text generation, which includes academic essays. This has sparked questions of how this technology will affect the education system, with students potentially able to auto-generate their coursework. However, a new app called GPT Zero might be the solution to this. Here is what GPT Zero means for universities in the wake of these concerns.
What is GPT Zero?
GPT Zero was designed by 22 year old Princeton student Edward Tian. It is an application made to analyse texts to determine whether they were generated by an AI. The app comes as a response to ethical questions regarding technologies such as ChatGPT being used to pass off AI work as one’s own, including in a classroom environment. It is currently in its beta version.
See also: What Does Pog Mean? The Slang And Gaming Definition Explained
The app works by measuring a text’s ‘perplexity’ — the verbal complexity and randomness that comes from work written by a human, as well as its ‘burstiness’ — the variations in sentence structure that people use. In other words, a work that is simple and familiar in its word choice and uses uniform sentences is a hallmark of something written by an AI, whereas something more complex and varied demonstrates a human writer.
How Could It Impact Universities?
GPT Zero, while still in development, has the potential to meet a need within schools and universities as AI technologies improve. It could be used to vet admissions essays to ensure that applicants have done the work themselves. It could also be used as standard procedure when coursework is submitted, similar to how plagiarism checkers are used currently. This would take some of the burden off of educators, so that they don’t have to make a judgment call themselves whether a piece of work seems like it was generated by an AI or not.
However, it is likely to take some time before this is fully implemented. GPT Zero has promising results identifying which texts are human made and which were written by an AI, but it does have some limitations. It requires at least ten sentences, so shorter texts can’t be analysed currently. It also has no claims of being able to always make its determination correctly. With widespread testing and fine-tuning to a good degree of accuracy, however, it may well prove to be an invaluable tool.
See also: How Do AI Generators Affect Artists?