Researchers at OpenAI have discovered a method to transfer adversarial robustness between different types of perturbations. This means that if a model is trained to be resilient against one type of attack, it can also withstand others. The team achieved this by using a technique called 'meta-learning', which allows the model to learn how to adapt to new situations. According to OpenAI, this breakthrough has implications for improving the security of AI systems. --- Why it matters: This matters because it could lead to more robust and secure AI models that can withstand various types of attacks, reducing the risk of AI systems being compromised or manipulated. Source: https://openai.com/index/transfer-of-adversarial-robustness-between-perturbation-types