OpenAI has a system that watermarks ChatGPT-created text and a tool to detect the watermark ready for almost a year, reports The Wall Street Journal. However, the company is divided internally over whether to release it. On one hand, it seems like the responsible thing to do: on the other, it could hurt something else.
The company’s watermarking is described as adjusting how the model predicts the most likely words and phrases that will follow the old ones, creating a detectable pattern. Offering a way to detect AI-written material could be a boon for teachers trying to stop students from using AI for their assignments.
The Journal reported that the company found watermarking didn’t affect the quality of its chatbot’s text output. In a survey the company said, “People worldwide supported the idea of an AI detection tool by a margin of four to one,” the Journal writes.
Also Read: Nvidia’s Next AI Chip Delayed Due To Design Flaw: Report
How Will This New Functionality Help Users?
After the Journal published the news, OpenAI confirmed it has worked on watermarking text in a blog post update today that was first spotted by TechCrunch. The company claimed its method is very accurate and resistant to “tampering, such as paraphrasing.” But it says techniques like rewording with another model make it “trivial to circumvention by bad actors.”
OpenAI also claims it’s concerned about the stigmatization of AI tools’ usefulness for non-native speakers. But it looks like the company is also worried that using watermarking could turn off surveyed ChatGPT users, almost 30 percent of whom evidently told the firm that they’d use the software less if watermarking was implemented.
Despite that, some employees have suggested that watermarking is effective. The Journal says some suggested trying methods that are “potentially less controversial among users but unproven.”
Also Read: iPhone 16 Pro Could Have Darker Black Titanium Color Option
OpenAI Rolling Out Her-Like Voice Mode For ChatGPT
OpenAI’s new advanced voice mode for ChatGPT is starting to roll out a small number of users who subscribe to ChatGPT Plus. The feature, which the company showcased at its GPT-4o launch event in May, was criticized for sounding similar to Hollywood actress Scarlett Johansson and was later delayed due to safety concerns.
During the OpenAI event, the new voice mode made a cameo with more capabilities compared to ChatGPT’s current voice mode. OpenAI employees were able to interrupt and ask the chatbot to tell a story in different ways, and the chatbot took their interruptions in stride to adjust its responses.
The advanced mode was set to release in alpha in June, but OpenAI delayed the rollout by one month in order to “reach our bar to launch.” As part of that delay, the company said it was “improving the model’s ability to detect and refuse certain content.”