Google launches SynthID Text technology for labelling AI-generated text
Google has unveiled SynthID Text, a technology that allows developers to add watermarks and discover text created by generative AI models. SynthID Text is available for download on the Hugging Face platform and in Google's updated Responsible GenAI Toolkit.
Here's What We Know
The technology works by changing the probability of token generation, which creates a unique pattern that serves as a watermark. This pattern helps determine whether the text was created by AI or other sources.
Today, we're open-sourcing our SynthID text watermarking tool through an updated Responsible Generative AI Toolkit.
- Google DeepMind (@GoogleDeepMind) October 23, 2024
Available freely to developers and businesses, it will help them identify their AI-generated content. ????
Find out more → https://t.co/n2aYoeJXqn pic.twitter.com/4uRKYaz57Y
Google claims that the quality, accuracy and speed of text generation does not suffer even if the text has been altered or translated. However, the technology has limitations, especially when dealing with short texts or answers to factual questions. In the future, it is possible to introduce mandatory watermarks for AI-generated content, as has already been done in China.
This innovation is important because, according to a report by the European Union Law Enforcement Agency, up to 90 per cent of online content could be created by AI by 2026, provoking new challenges for law enforcement agencies in relation to misinformation, propaganda, fraud and deception. The study shows that already today around 60 per cent of all texts on the internet can be produced by artificial intelligence, driven by the widespread use of AI translators.
Source: @GoogleDeepMind, Tech Crunch
