English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

A growing ecosystem of large, open-source foundation models has reduced thelabeled data and technical expertise necessary to apply machine learning tomany new problems. Yet foundation models pose a clear dual-use risk,indiscriminately reducing the costs of building both harmful and beneficialmachine learning systems. Policy tools such as restricted model access andexport controls are the primary methods currently used to mitigate suchdual-use risks. In this work, we review potential safe-release strategies andargue that both policymakers and AI researchers would benefit fromfundamentally new technologies enabling more precise control over thedownstream usage of open-source foundation models. We propose one suchapproach: the task blocking paradigm, in which foundation models are trainedwith an additional mechanism to impede adaptation to harmful tasks withoutsacrificing performance on desirable tasks. We call the resulting modelsself-destructing models, inspired by mechanisms that prevent adversaries fromusing tools for harmful purposes. We present an algorithm for trainingself-destructing models leveraging techniques from meta-learning andadversarial learning, which we call meta-learned adversarial censoring (MLAC).In a small-scale experiment, we show MLAC can largely prevent a BERT-stylemodel from being re-purposed to perform gender identification without harmingthe model's ability to perform profession classification.

Self-Destructing Models: Increasing the Costs of Harmful Dual Uses of Foundation Models