Study warns that misaligned AI models can spread harmful behaviours
The article shows how, after fine-tuning on a very specific task (e.g., writing deliberately unsafe code), the model begins to give toxic or harmful ...
Science Media Centre España 2
Covers: AI Safety & Existential Risk
More biosecurity due to dangerous proteins with AI - Science Media Centre España
Artificial intelligence ( AI )-assisted protein engineering is enabling advances in the design of new molecules, but it also poses biosecurity ...