The world of technology is constantly evolving with the development of artificial intelligence. Google is working on a new digital watermark technology called SynthID TexT to distinguish texts generated by artificial intelligence. This technology will be accessible via the Hugging Face AI platform and Responsible GenAI Toolkit and will be offered as open source. This move by Google will provide developers and businesses with a better opportunity to understand content generated by artificial intelligence.
How SynthID TexT Works
SynthID TexT operates using the principle of token distribution by artificial intelligence models. This technology analyzes the probabilities of word formation in texts to create a pattern and marks the text according to this pattern. This makes it easy to detect content produced by artificial intelligence.
However, Google also notes that this technology has some limitations. In particular, short texts, articles translated to another language or rewritten, and knowledge-focused texts, SynthID TexT may not perform as expected. Nevertheless, overcoming these challenges is targeted as the technology is developed.
Additionally, Google has already integrated this technology into Gemini models. Marking artificial intelligence outputs has become legally mandatory in regions like China and California. With similar laws being adopted in other regions, the use of AI marking technologies is expected to increase. These developments are seen as an important step in managing and regulating AI-generated content.