Over the past decade, neural networks have been a revolution in the computing landscape. However, the energy demands of running deep learning models have become unsustainable on the current digital hardware. Therefore, as digital computing approaches its energy efficiency limits, especially for AI workloads, neuromorphic and analog computing have re-emerged as promising alternatives. Neuromorphic systems mimic how biological brains process information by leveraging parallelism, asynchronous signal propagation, and unified memory-computation units to optimize energy consumption. Despite neuromorphic chips can be used to speed up existing AI models, it is essential to design to develop new models and algorithms that can embrace and exploit these characteristics. On the other hand, analog computing leverages the continuous dynamics of physical systems, such as electrical currents or material properties, to perform computation at significantly lower power and latency than traditional digital systems. This makes analog approaches especially compelling for edge AI applications and energy-constrained settings. Despite their promises, both neuromorphic and analog computing raise key research questions, such as designing algorithms that are robust to the inherent noise of analog substrates, understanding the levels of precision and reliability for different AI tasks, and co-designing models and circuits to balance benefits with scalability and programmability. This mini-symposium explores brain-inspired and analog computing paradigms that promise transformative advances in energy-efficient AI. We discuss both theoretical foundations and practical applications of neuromorphic and analog computing, highlighting key challenges, recent breakthroughs, and future directions. The event aims to bring together machine learning researchers and hardware engineers to to promote discussion on the design of new generation AI systems for non-traditional and energy-efficient hardware platforms.