Название: AI-Processor Electronics: Basic technology of artificial intelligence
Автор: Vinod Kumar Khanna
Издательство: IOP Publishing
Год: 2025
Страниц: 399
Язык: английский
Формат: pdf (true), epub
Размер: 33.9 MB

The revolutionary impact of advances in Artificial Intelligence (AI) on consumer and industrial electronics has significantly increased the demand for specialized processing units for AI workloads. AI-Processor Electronics: Basic technology of Artificial Intelligence explores the principles and technology behind these processors, which perform fast, energy-efficient data processing and complex computations. The book begins with an overview of AI, Machine Learning, and Deep Learning, and a brief introduction to digital computer electronics. It highlights the limitations of traditional CPUs in modern AI, discussing parallel computing and AI-optimized hardware that are specially tailored and adapted to meet AI requirements. The book then covers various AI processors, including graphical, tensor, neural, convolutional neural network, vision, sparse neural network, graph analytics, associative memory, and quantum processing units.

High-performance AI processors deployed in data centers or at the edge can help unlock a world of new opportunities for businesses and industries. These chips have been the enablers of major AI advancements. Such leading-edge, specialized AI chips are essential for cost-effective AI implementation on a large scale. With ever-increasing sophistication of AI algorithms, the need for higher processing power, speed and efficiency has grown rapidly, raising the expectations form these devices for autonomous, smart processes in robotics, Internet of Things, computer vision, voice recognition and natural language understanding and many more fields.

The book apprises the reader about the type of challenges faced in AI computing. After a quick revision of the computing applications of digital electronics in a nutshell, the operation of CPU is explained and the versatility advantages vis-à-vis inadequacies of the general-purpose processors for handling AI tasks are pointed out. Then the ubiquitously prevalent parallel processing architecture in AI chips is presented followed by highlighting the approaches that must be incorporated for efficient computing keeping in view the deceleration of Moore’s law. A diversity of specialized processors is successively described such as the GPU, the TPU, the NPU, the CNN processor and VPU, the sparse NN processor, the graph processor and the in-memory processor. Finally, the rudimentary principles of quantum computing are introduced. Quantum computing devices, logic gates and circuits are dealt with and quantum processing unit (QPU) operation is summarized.

All in all, the book sketches a broad all-inclusive perspective of electronics for AI processing units, extending from classical to quantum-mechanical computing. It elucidates the various optimization approaches for AI processors including the use of parallel processing architecture in concurrently performing multiple calculations across data streams, besides low-precision computing, domain-specific architecture and the neural network sparsity concept. It offers a compendious exposition of the principles and technology of processors capable of performing fast and energy-efficient large-scale data processing and complex computations.

A large number of semiconductor chip manufactures are working on AI processors and a vast spectrum of processors are available in the market. It is a highly competitive and fast-developing field progressing by leaps and bounds. As new chips are introduced, older versions go into obsolescence and oblivion. It is well-nigh impossible to attempt an exhaustive treatment of the subject.

The book will help the reader in understanding the core foundational principles of AI processors, and their enabling fabrication technologies, and getting glimpses of the scenario. The book will be of immense value to graduate/post graduate and PhD students in electronics and computer engineering and related disciplines, and equally relevant to the practicing engineers and scientists engaged in this field.

Key features:

• Provides a review of the physics of operation of a diverse variety of representative AI processors, and the key concepts and innovations underlying their development.
• Covers the underlying principles, technologies, opportunities and future prospects.
• Highlights the inadequacies of general-purpose processors in meeting the computationally intensive needs of Artificial Intelligence workloads.
• Elucidates the use of parallel processing architecture in simultaneously performing multiple calculations across data streams.
• Explores the concepts of neuromorphic and quantum computing to address the growing demands of big data in AI.