As the rise of ChatGPT creates worries about plagiarism, Israeli firm Copyleaks can spot texts that come from AI tools. It does not only spot texts from the popular ChatGPT but also texts from other generative AI tools. The company is now showing its ability to academia in the U.S. to help verify the originality of students’ work.
A simple test reveals the generic nature of texts coming out of corporate top-level officials. In this test, ChatGPT was packed to draft a message coming from Google’s CEO warning staff of layoffs. This test was conducted on the same day that Google laid off 12,000 staff. The letter from ChatGPT and the real letter from Google’s CEO, Sundar Pichai was placed side by side. Random people were then asked to choose which of the letters was written by a human. The result of the tests shows that all the participants choose the AI text as the one that came from a human. However, an AI tool from an Israeli startup swiftly spotted the letter from ChatGPT. In its remarks, it said, “Written by AI with a high probability”. The Israeli startup, Copyleaks has offices in New York and an R&D center in Kiryat Shmona.
Alon Yamin, the company’s founder and CEO, claims that he is not surprised that the tool can spot AI text. According to him, this new tool has a copy-prevention system which helps it to identify AI texts. Also, Yamin claims that this tool is already revealing how much people already rely on AI texts.
In an interview at Copyleaks New York office, Yamin told Calcalist
“We have access to a lot of content from students and researchers… In the last few weeks, we have started to activate AI detection on the content, to know the percentage of students who use artificial intelligence to write content. In the last three weeks, over 10% of the content submitted to the system, which is hundreds of thousands of documents, included text created by AI, and that’s when ChatGPT just came out. The data will continue to rise. We were very surprised by this number.”
A couple of AI tools (ChatGPT is chief of them) can produce pretty decent high-level texts. ChatGPT for instance has been proven to be decent enough to pass professional exams in medicine, accounting and even some advanced courses. In some tests, ChatGPT scores no less than a B grade and this is a huge concern in the academic world. However, for Yamin, Copyleaks offers the solution to this issue. He also believes that the tool will take care of this crisis.
Gizchina News of the week
“Students need to know how to write content, this is an important skill that will not disappear from the world, but there is a process of deciphering how to work with these tools. Everything is very, very new.”
According to Yamin, Copyleaks was founded 8 years ago by himself and Yehonatan Bitton, VP of R&D. The focus of the company is the analysis of text for AI tech. The company simply analysis AI texts, knowing its source, originality, tone and other parameters.
We were prepared for ChatGPT
According to Yamin, the company has been on a tech to reduce the risk of AI tools like ChatGPT. He claims that they saw the emergence of ChatGPT months ago before it hit the public. Copyleaks saw the need to detect texts that comes from AI and this is all its tech does. He adds that humans may never be able to separate AI texts from texts written by other humans. Thus, only AI tools can identify AI texts. He adds
“It may be difficult for people to recognize the difference, but at the end of the day, an AI system writes in a different way that looks different from a statistical point of view… In the end, it’s about the text, even if it’s created by AI, and we’re constantly working on analyzing text content…”
How does Copyleaks AI text detection tool work?
In explaining how the system works Yamin says that a knock on the door may sound regular to many people. However, those who know Morse Code will be able to tell the meaning of the knock. He claims that Copyleaks system is an AI tool that can tell the difference between AI and non-AI texts. Since the system understands the origin of AI texts, it can tell the texts that have AI origin.
Since the launch of ChatGPT, a slew of programs claims to be able to detect text written by it. OpenAI, the chatbot’s creator, also intends to release its own ID tool for its texts. However, Copyleaks still believes that it has an advantage. Yamin said