ChatGPT achieved a huge boom and wide spread in a short period of time, but this called at the same time a counter wave calling for the detection of counterfeiting that could occur due to these artificial intelligence techniques, which eventually led to the emergence of new tools to confront it, on top of which is a new application called GPTZero, which he developed A young man of only 22 years old.
Edward Tian, a 22-year-old undergraduate student studying computer science and journalism at Princeton University, is the creator of GPTZero, a free and publicly available tool that can detect whether text is written by a human or a machine.
GPTZero: A new AI content detection app
Tian put the code behind the app earlier in January, after returning home to Toronto, Canada, during vacation. Tian initially thought only a few dozen people would try the tool, but so far, GPTZero has received quite a bit of interest from outlets. The main news, in addition to the application's website, which has had more than 7 million visits since its launch.
Tian, who is currently doing his thesis research on the discovery of artificial intelligence, wanted to create something that would help a wide range of people—teachers, college admissions officers, or anyone reading an article online—know what they might be dealing with.
A new AI content detection app is GPTZero.
And the rise of AI discovery tools like GPTZero comes at a time when AI seems to be making controversial waves in almost every field, be it academia, technology or the art world, in which models like ChatGPT - a conversational language paradigm developed by OpenAI - Powered by an unimaginably huge amount of text being scraped from the Internet, ChatGPT is built on top of the company's family of powerful GPT-3 models, mapped to over 175 million parameters - an unprecedented number at the time of its initial launch in 2020. .
How does GPTZero
GPTZero works by analyzing text for some indicators, such as what Tian calls "perplexity" and "impulsivity". Confusion refers to the complexity or randomness of text. If GPTZero is "confused" by the text it is evaluating, it is more complex, and therefore, than Likely human-composed, impulsivity refers to patterns of variation in sentence structure, which can indicate whether or not a written text is machine-generated.
Tian is now working to further improve the accuracy of the app. In addition, he is in discussions with various school boards and scholarships about using a newer corporate version of the GPTZeroX model. However, Tian still plans to keep the online copy-and-paste version of the app free for widespread public use. wide.
While Tian's development of GPTZero stems from his interest in the responsible use of AI, Tian maintains that he does not want to ban the use of AI indiscriminately. to become more widespread in the future.
OpenAI attempts
On the other hand, ChatGPT developer OpenAI has introduced a tool to identify text generated by ChatGPT. The makers of the world's most famous chatbot have introduced a solution to help people distinguish between human-generated text and a bot.
OpenAI, the company behind ChatGPT and creator of DALL-E text-to-image, said it had "trained the classifier to distinguish between text written by a human and text written by an AI".
But there's a catch: it's not quite reliable yet. In testing so far, 26% of texts written by AI were flagged as "likely AI-written," while typed text was incorrectly categorized as typed. By AI 9% of the time, the tool has proven to be most effective on batches of texts over 1,000 words in length, but even then the results are quite questionable.