ChatGPT"s developer OpenAI has confirmed that it has a tool ready that will help expose cheating students. Reportedly, Open AI is sitting on a text watermarking tool that has a high degree of accuracy and can detect essays or text generated using ChatGPT.
In an update to their official blog that was published on May 4, Open AI said, "Our teams have developed a text watermarking method that we continue to consider as we research alternatives." The Wall Street Journal brought light to this new OpenAI"s text watermarking tool, but the debate over whether the tool should be released or not is keeping it shelved.
OpenAI further added in their official note that:
While it has been highly accurate and even effective against localized tampering, such as paraphrasing, it is less robust against globalized tampering; like using translation systems, rewording with another generative model, or asking the model to insert a special character in between every word and then deleting that character - making it trivial to circumvention by bad actors.
OpenAI is looking at text watermarking as an additional tool over solutions such as classifiers and metadata. The company supposedly gave its side of the reason why the text formatting tool hasn"t seen the light of the day. OpenAI added that it could, "disproportionately impact some groups. For example, it could stigmatize the use of AI as a useful writing tool for non-native English speakers."
For now, the company is evaluating the risks involved in releasing the text formatting tool, since it will likely impact a broader ecosystem beyond OpenAI. OpenAI further noted that it is prioritizing releasing the authentication tools for audiovisual content.
Recently, OpenAI announced GPT-4o Long Output with support for 64K output tokens. Google also slashed the pricing for Gemini 1.5 Flash, giving rise to a price war as the new prices make Gemini 1.5 Flash nearly 50% cheaper than OpenAI"s GPT-4o mini.