Google will now watermark AI text

Google has created a tool called SynthID Text. This tool can detect if something was written using AI or not. The company has recently moved to make it available to the general public, including both business owners and developers. Not only can it identify AI text, but it can add a watermark so that anyone else who is reading the text will also know that it was not written by a human.

This is very important regarding accuracy. Generative AI systems use “tokens.” These tokens can be a word or even just one single character. The AI can analyze which token most commonly comes after another, and this is how it can generate blocks of text. It does this at an incredible speed so that the user isn’t precisely aware of what’s going on behind the scenes, but these building blocks allow the system to create entire documents.

But it is important for readers to know if the text on their screen was written by a human or if it is essentially just a guess from a generative AI model. New tech tools will simply make the process more transparent so that users will be more likely to know where they are getting their information from and whether or not they can trust it.

Does SynthID Text always work?

This is all relatively new technology, so it is certainly not perfect. Google notes that it doesn’t always work well with translations or with short text. In some cases, there is not much room for flexibility, so the program may not know if the text was AI-generated or not.

For example, if an AI system was asked to recite a specific poem, it would probably provide a word-for-word copy. SynthID Text could check this, but may not be able to determine if it was generated or not because the text would be identical.

Concerningly, Google also notes that SynthID Text doesn’t work as well when users are asking factual questions. Say that someone asked an AI for the capital of New York and then put that information on a webpage. Since it is a short and factual answer, Google may not be able to determine if the person wrote it or if they used a generative AI model.

The complexities of AI

AI is changing the Internet, and it’s raising a lot of questions about accuracy and transparency. Google is certainly attempting to address this with SynthID Text, but it does show how complex the situation can become.

Here at Content Customs, we always stay in touch with the newest technological developments, and we have decades of experience creating online content. You can trust us to provide content that is accurate and relevant, and we can assist with everything from search engine optimization to online marketing. Just contact us today to learn more.