The team at artificial intelligence company OpenAI has developed a method for watermarking text and will continue to consider it as it explores alternatives.
Although it is highly accurate and even effective at resisting local tampering such as paraphrases, it is less resistant to global tampering; for example, using a translation system, rewriting it with another generative model, or requiring the model to insert a special character between each word and then delete it, making it easy for bad actors to circumvent it.
and may have a disproportionate impact on groups such as non-English speakers. The company is discussing whether to actually release the tool. OpenAI shut down its previous AI text detector last year, citing “low accuracy.”
OpenAI is developing a novel text watermarking technology to identify and detect text generated by ChatGPT. This technology is achieved by embedding invisible watermarks in the text, which is achieved by making minor changes to the way ChatGPT selects words, essentially creating an invisible watermark in the text that can later be detected by a separate tool.
However, there are still many challenges. For example, in the face of text translation, rewriting, or insertion of special characters, the recognition effect of watermarks may be greatly compromised. At the same time, the technology may have an unfair impact on users who do not speak English as their native language. Given these complexities, OpenAI said it will continue its research cautiously and focus more on the development of audio-visual content authentication tools.
OpenAI has developed a 99.9% effective ChatGPT text watermarking system, but they hesitated to release it due to potential user backlash.
Although a survey showed strong support for AI Detection Tools, OpenAI is concerned that watermarks may prevent users from using them, with nearly 30% saying they will reduce their use of ChatGPT. There are also internal concerns that watermarks can be easily circumvented. Despite these challenges, OpenAI is exploring other, less controversial ways to detect AI-generated text.
If you want to learn more, you can click on the link below the video.
Thank you for watching this video. If you like it, please subscribe and like it. thank
Oil tubing: