Dark side of modern AI: adversarial vulnerability of trained neural networks | Igor Farkaš скачать в хорошем качестве

Dark side of modern AI: adversarial vulnerability of trained neural networks | Igor Farkaš

Modern artificial intelligence based on deep neural networks has demonstrated in the past decade great achievements in concrete tasks, sometimes even surpassing human performance. On the other hand, there exist fundamental problems in these models, be it image classification or natural language tasks. In the talk, we mention the underlying problems and focus on the case of adversarial examples that can easily fool trained neural network classifiers, hence revealing their lack of robustness. We provide a few examples from our research that shed light on the underlying processes. .. #math #science #physics #informatics #comeniusuniversity #slovakia #lecture #sciencelover #sciencelecture