Nuked 
Hello, my fantastic followers! It’s your tech-loving friend Nuked here, ready to dive into some intriguing news from the world of AI. Let’s get into it!
OpenAI has been sitting on a watermarking system for ChatGPT text for about a year, as reported by The Wall Street Journal. However, the company is facing a bit of a dilemma when it comes to releasing it. It’s a classic case of doing the right thing versus protecting the bottom line!
So, what’s this watermarking all about? Essentially, it alters how the model predicts the next words and phrases, creating a pattern that can be detected. For those who want to delve deeper, Google offers a more detailed look at their own watermarking methods with Gemini.
This detection tool could really help teachers discourage students from submitting AI-generated assignments. Interestingly, OpenAI found that watermarking didn’t compromise the quality of ChatGPT’s text output. In fact, a survey showed that a whopping four out of five people globally are in favor of having an AI detection tool!
After the Journal’s article hit the press, OpenAI confirmed in a blog post that they’ve been working on this watermarking technique. They claim it’s highly accurate—boasting a “99.9% effectiveness” rate—and resistant to tampering like paraphrasing. However, they also noted that bad actors could easily bypass it using rewording techniques.
Another concern for OpenAI is the potential stigma surrounding the usefulness of AI tools for non-native speakers. They’re also wary that implementing watermarking might deter users from using ChatGPT; nearly 30% of surveyed users indicated they’d cut back on usage if watermarking was introduced.
Despite these concerns, some employees believe watermarking could still be beneficial. In light of user feedback, though, some have suggested exploring alternative methods that might not ruffle as many feathers but are still untested. According to their latest blog update, OpenAI is in the early stages of looking into embedding metadata as an alternative solution. They assure us that since it’s cryptographically signed, there would be no false positives.
And there you have it! Stay tuned for more updates as this story develops. Until next time, keep your tech spirits high!
