Bridging the Digital Divide - Generative AI at the Edge

Explore how edge computing can democratize generative AI access and reduce the global digital divide in this 32-minute conference talk by IEEE Fellow Danilo Pau from STMicroelectronics. Discover why current AI development's centralized approach creates barriers to widespread adoption and learn practical solutions for deploying generative AI on low-power edge devices. Examine real-world applications including intelligent thermostats with natural language capabilities, TinyML-powered vision systems, style transfer on microcontrollers, and edge-based language models and visual question answering systems. Understand how to design natural human-machine interfaces that can reason locally without cloud dependency, with particular focus on audio processing as an accessible entry point for edge AI implementation. Gain insights into energy-efficient strategies for running generative AI models on resource-constrained devices, making advanced AI capabilities available to users regardless of their computational resources or internet connectivity. Learn how this decentralized approach to AI deployment can create more inclusive technology solutions while reducing environmental impact and improving privacy through local processing.