Attackers can elicit ‘toxic behavior’ from AI translation systems, study finds [VentureBeat]

In a new study, researchers demonstrate a flaw in neural machine translation systems that can prompt them to generate toxic outputs.

AI,Big Data,Dev,VB Home Page,adversarial attack,ai,artificial intelligence,Attack,category-/Reference/Language Resources,flaw,machine learning,neural machine translation,NMT,research,Security,study,translation

adversarial attack