Smart Software: It May Never Forget
November 13, 2024
A recent paper challenges the big dogs of AI, asking, “Does Your LLM Truly Unlearn? An Embarrassingly Simple Approach to Recover Unlearned Knowledge.” The study was performed by a team of researchers from Penn State, Harvard, and Amazon and published on research platform arXiv. True or false, it is a nifty poke in the eye for the likes of OpenAI, Google, Meta, and Microsoft, who may have overlooked the obvious. The abstract explains:
“Large language models (LLMs) have shown remarkable proficiency in generating text, benefiting from extensive training on vast textual corpora. However, LLMs may also acquire unwanted behaviors from the diverse and sensitive nature of their training data, which can include copyrighted and private content. Machine unlearning has been introduced as a viable solution to remove the influence of such problematic content without the need for costly and time-consuming retraining. This process aims to erase specific knowledge from LLMs while preserving as much model utility as possible.”
But AI firms may be fooling themselves about this method. We learn:
“Despite the effectiveness of current unlearning methods, little attention has been given to whether existing unlearning methods for LLMs truly achieve forgetting or merely hide the knowledge, which current unlearning benchmarks fail to detect. This paper reveals that applying quantization to models that have undergone unlearning can restore the ‘forgotten’ information.”
Oops. The team found as much as 83% of data thought forgotten was still there, lurking in the shadows. The paper offers a explanation for the problem and suggestions to mitigate it. The abstract concludes:
“Altogether, our study underscores a major failure in existing unlearning methods for LLMs, strongly advocating for more comprehensive and robust strategies to ensure authentic unlearning without compromising model utility.”
See the paper for all the technical details. Will the big tech firms take the researchers’ advice and improve their products? Or will they continue letting their investors and marketing departments lead them by the nose?
Cynthia Murrell, November 13, 2024
Comments
Got something to say?