Trustworthy AI in the Cloud

Nowadays, AI applications are becoming extremely popular in our every-day life as well as for the industry. Therefore, ensuring the trustworthiness of AI hardware is crucial, especially for edge applications in safety-critical systems. How-ever, recent incidents involving major tech companies such as Google, Meta, and Alibaba have revealed that even cloud-based data center hardware can experience failures leading to Silent Data Corruptions (SDCs), also called Silent Data Errors (SDEs). This chapter delves into the implications of such failures on AI workloads, both during training and inference, and explores methodologies for efficiently detect-ing SDCs or SDEs through dedicated hardware monitors.