DeepMind launches tool that detects the use of artificial intelligence in content such as text and images
I launched recently AI Generated Text Detector – Synth ID Text Synth ID-Text, from Google-owned DeepMind, to provide a text watermarking system, and is the first tool to be tested on 20 million verification queries.
Google has announced that it has already integrated the new watermarking system. Synth ID-Text in its Gemini chatbot. It has also open-sourced the tool and made it available to developers and businesses, allowing them to use the tool to determine whether text output came from their own large language models (LLMs), the AI systems that power their chatbots.
This comes after content verification data has been a hot topic, especially for images and videos, to combat deepfakes. Detecting the source of the text is a much more difficult dilemma, as the text can be easily changed to hide or remove the watermark by removing the formatting.
The tool is intended to counter the flood of AI-generated content, including images, videos, and text, that has been creeping into everything from news feeds and research papers to emails, necessitating the development of countermeasures and steps to uncover the source of this content.
Some companies offer services to identify AI-generated text by analyzing the material, while others say their tools will make AI-generated text look “human” and make the use of AI to generate it undetectable.
There are currently some tools available that provide the service of detecting the sources of content suspected of being generated by artificial intelligence, such as:
- TraceGPT is a tool for text accuracy and analysis.
- Winston AI to detect the source of images and texts.
- Hive is a free tool for detecting the source of text and images.
- GPTZero has the ability to analyze written text and its sources.
- Originality.ai is a general content detection tool.
- Smodin is a low-cost tool for uncovering the source of content.
