The Janus Interface - Privacy Risks in Fine-Tuning Large Language Models

Explore a groundbreaking research presentation on the Janus attack, a novel method exploiting the fine-tuning interface in large language models to recover forgotten personally identifiable information from pre-training data. Delve into the work of Xiaoyi Chen and co-authors from Indiana University as they unveil the privacy risks amplified by fine-tuning in LLMs. Gain insights into the implications of this research for AI security and data protection. Learn about the latest developments in AI research and industry trends through additional resources, including The Deep Dive newsletter and Unify's blog. Connect with the AI community through various social media platforms and engage in discussions about the future of machine learning and deep learning.