§ feed · storyline
Robust adversarial inputs
OpenAI publishes adversarial images that reliably fool neural network classifiers across varied scales and perspectives, countering claims that multi-angle capture makes self-driving car vision systems hard to deceive.
We’ve created images that reliably fool neural network classifiers when viewed from varied scales and perspectives. This challenges a claim from last week that self-driving cars would be hard to trick maliciously since they capture images from multiple scales, angles, perspectives, and the like.
§ sources1 publication · timeline below
- openai.comRobust adversarial inputsprimary