SynthID-Text:Watermarking AI-generated text content
Large language models (LLMs) have enabled the generation of high-quality synthetic text, often indistinguishable from human-written content, at a scale that can markedly affect the nature of the information ecosystem.
Watermarking can help identify synthetic text and limit accidental or deliberate misuse, but has not been adopted in production systems owing to stringent quality, detectability and computational efficiency requirements.
In this talk, the leading author of the work, Sumanth Dathathri, a research scientist at Google DeepMind, will introduce SynthID-Text, a production-ready text watermarking system designed to address these requirements.
SynthID-Text seamlessly integrates with LLMs, preserving the quality of generated text while enabling accurate detection with minimal latency.
Our guest will explore its technical intricacies, including its compatibility with advanced techniques like speculative sampling, crucial for enhancing the efficiency of production systems. Through rigorous empirical evaluations across diverse LLMs and standard benchmarks, complemented by human side-by-side ratings, we demonstrate that SynthID-Text does not hinder LLM capabilities.
Sumanth will also share insights from a live experiment involving nearly 20 million Gemini responses, showcasing how SynthID-Text maintains text quality based on direct user feedback.
Join the BuzzRobot Slack to connect with the community