ABSTRACT
This chapter describes emerging hardware platforms, such as graphics processing units (GPUs) and Intel's high-performance Xeon processors, and industry standard software frameworks, such as Hadoop, TensorFlow, and Spark and also discusses the opportunities for fixed function acceleration of AI workloads. It presents an overview of AI system architecture, including statistical machine learning and deep-learning principles along with key use cases. The chapter introduces the art of data science and end-to-end workflows and shows machine- and deep-learning software frameworks, libraries, and primitives. It describes various hardware capabilities and their architectural fit for key use cases and explores how AI software interacts with hardware for key workloads. The chapter also summarizes new innovations and future research work in this area. Artificial intelligence software frameworks rely on software libraries from silicon vendors, such as Intel and Nvidia, to access instruction set architectures without having to program at assembly level.
