publish time

07/08/2024

author name Arab Times

publish time

07/08/2024

OpenAI's AI detection tool creates invisible watermarks to combat cheating.

NEW YORK, Aug 7: When OpenAI launched ChatGPT in late 2022, concerns arose among educators about the potential for students to use the platform to cheat on assignments and tests. In response, various companies have introduced AI detection tools, though their effectiveness has been questioned.

OpenAI has now unveiled its own solution, designed to detect when ChatGPT is used to generate text. According to The Washington Post, the new technology boasts a 99.9% accuracy rate. It works by predicting the next word or phrase, known as a “token,” in a sentence. The AI-detection tool subtly alters these tokens, creating an invisible watermark that can be detected by the tool but remains undetectable to the human eye.

Although the technology has been ready for nearly a year, OpenAI has yet to launch it publicly. Internal reactions to the tool have been mixed. Some worry that its release might alienate a segment of ChatGPT's user base, while others argue that it aligns with OpenAI's commitment to transparency.

An OpenAI spokeswoman expressed concerns that the tool could disproportionately impact non-native English speakers. Despite this, some key employees believe the benefits of the technology outweigh the risks and advocate for its release.

Generative AI tools like ChatGPT and Gemini have revolutionized how students and educators interact with technology, offering capabilities such as answering questions and drafting research papers from minimal prompts. However, the watermarks created by AI tools can sometimes be removed using techniques like adding and removing emojis or translating the text into other languages and back.

Additionally, there are concerns about the practical application of the AI detection tool. If usage is too low, the tool might become redundant. Conversely, widespread use could lead to the watermarking technique becoming known and circumvented.

The future will reveal whether OpenAI will move forward with this detection tool, potentially impacting both its own operations and the broader educational landscape.