Just over a year after its announcement in August 2023, the technology giant, Google has launched an open source version of SynthID, its artificial intelligence (AI) tool that will allow developers, companies and anyone interested to identify and place watermarks in texts generated with linguistic models. (LLM).
It was through a publication on the X account of Google DeepMind, the division in charge of the study and development of artificial intelligence, that the company announced the launch of this tool, which, although it is not the first attempt by companies to combat misinformation and the allocation of material, Yes, we can say that it is the first large-scale attempt in recent years.
With the growing use of artificial intelligence and the increasingly recurrent options at our fingertips, it is difficult to find measures that counteract the misuse of these language models, as well as help developers perfect the existing ones. However, although it is usually easy to add watermarks to images, videos or audio, in texts any modification can be detrimental to its understanding.
READ: Learn about the new playlist features on Spotify
It is also important to note that although SynthID will allow texts created with LLM to be marked, these will not be visible to the naked eye and we will only be able to observe them through specialized identification software.
Ahead of the open source release this morning, Google has carried out tests with its language model, Google Gemini, in which positive results have been obtained, highlighting the precision of SynthID when it comes to long texts. However, the tests have not revealed the level of resistance that the watermarks have to any removal attempt, but it is expected that in the coming days we will have more details on the uses of this tool.
JM
Themes
- Artificial intelligence
