k.Work
WorkOpenAI Blog
KeyRank 55Adversarial attacks on neural network policies
OpenAI publishes research on adversarial vulnerabilities in neural network-based policies, surfacing a critical safety and robustness gap that affects deployment of AI systems in real-world decision-making contexts. This is foundational safety/governance research relevant to boards assessing AI risk.
Why it ranks · · OpenAI research on adversarial attacks against neural network policies · February 2017
Read full story