Intelligence Artificielle

Researchers Develop New Technique to Wipe Dangerous Knowledge From AI Systems

study published Tuesday provides a newly-developed way to measure whether an AI model contains potentially hazardous knowledge, along with a technique for removing the knowledge from an AI system while leaving the rest of the model relatively intact. Together, the findings could help prevent AI models from being used to carry out cyberattacks and deploy bioweapons.

The study was conducted by researchers from Scale AI, an AI training data provider, and the Center for AI Safety, a nonprofit, along with a consortium of more than 20 experts in biosecurity, chemical weapons, and cybersecurity. The subject matter experts generated a set of questions that, taken together, could assess whether an AI model can assist in efforts to create and deploy weapons of mass destruction. The researchers from the Center for AI Safety, building on previous work that helps to understand how AI models represent concepts, developed the “mind wipe” technique.

Dan Hendrycks, executive director at the Center for AI Safety, says that the “unlearning” technique represents a significant advance on previous safety measures, and that he hopes it will be “ubiquitous practice for unlearning methods to be present in models of the future.”

As the AI industry continues to make rapid progress, safety is top of mind for world leaders. U.S. President Joe Biden’s AI Executive Order, signed in October 2023, directs officials to take steps to “understand and mitigate the risk of AI being misused to assist in the development or use of [chemical, biological, radiological, or nuclear] threats,” and to mitigate cybersecurity risks posed by AI.

Source

Veille-cyber

Share
Published by
Veille-cyber

Recent Posts

Les 7 menaces cyber les plus fréquentes en entreprise

Introduction La cybersécurité est devenue une priorité stratégique pour toutes les entreprises, grandes ou petites.…

3 jours ago

Cybersécurité : Vers une montée en compétence des établissements de santé grâce aux exercices de crise

Cybersécurité : les établissements de santé renforcent leur défense grâce aux exercices de crise Face…

2 semaines ago

Règlement DORA : implications contractuelles pour les entités financières et les prestataires informatiques

La transformation numérique du secteur financier n'a pas que du bon : elle augmente aussi…

2 semaines ago

L’IA : opportunité ou menace ? Les DSI de la finance s’interrogent

L'IA : opportunité ou menace ? Les DSI de la finance s'interrogent Alors que l'intelligence…

2 semaines ago

Telegram menace de quitter la France : le chiffrement de bout en bout en ligne de mire

Telegram envisage de quitter la France : le chiffrement de bout en bout au cœur…

2 semaines ago

Sécurité des identités : un pilier essentiel pour la conformité au règlement DORA dans le secteur financier

Sécurité des identités : un pilier essentiel pour la conformité au règlement DORA dans le…

2 semaines ago

This website uses cookies.