Designing Energy-Efficient Intelligent Systems Using Probabilistic Computing and p-Bits

Review Article

Abstract

The rapid expansion of artificial intelligence (AI), Internet of Things (IoT), edge computing, and autonomous systems has significantly increased the demand for energy-efficient computing technologies capable of supporting intelligent decision-making under stringent power constraints. Conventional deterministic computing architectures based on CMOS technology and the von Neumann model face growing challenges in terms of energy consumption, memory bottlenecks, and computational scalability. Probabilistic computing has emerged as a promising alternative paradigm by exploiting controlled stochastic behavior rather than deterministic binary operations. Central to this approach are probabilistic bits (p-bits), which fluctuate between binary states with tunable probabilities, enabling efficient implementation of optimization, inference, machine learning, and combinatorial computing tasks. Unlike quantum bits, p-bits operate at room temperature using existing semiconductor technologies, making them practical for near-term deployment. This review presents the fundamental principles of probabilistic computing and p-bit architectures, emphasizing their role in developing ultra-low-power intelligent systems. The paper discusses the theoretical foundations of probabilistic computation, stochastic information processing, p-bit devices, probabilistic circuits, and hybrid probabilistic-classical computing frameworks. By examining these foundational concepts, the review provides a comprehensive understanding of how probabilistic computing offers a scalable and energy-efficient alternative for future intelligent computing systems operating in resource-constrained environments.