A Unified Memory Perspective for Probabilistic Trustworthy AI

Researchers present a unified framework for probabilistic AI computation that treats deterministic and stochastic data access under a common perspective. The study identifies memory systems as performance bottlenecks in trustworthy AI and proposes compute-in-memory approaches to address scalability challenges.

• →Memory systems become performance bottlenecks in probabilistic AI workloads rather than arithmetic units.
• →A unified framework treats deterministic data access as a limiting case of stochastic sampling.
• →Increasing stochastic demand reduces effective data-access efficiency and can drive entropy-limited operation.
• →Compute-in-memory approaches that integrate sampling with memory access show promise for scalable trustworthy AI hardware.
• →The research defines evaluation criteria including unified operation, distribution programmability, and parallel compatibility for probabilistic computing systems.