World Economic Forum Pushes Edge AI

The World Economic Forum's push for edge AI in SMEs centers on moving from cloud-based systems to real-time data processing on or near industrial machines. This shift reduces latency and dependency on constant connectivity, which is critical for SMEs in resource-limited areas. The WEF's "AI Playbook for India's MSMEs" outlines a cluster-based deployment model to encourage shared learning and infrastructure among these smaller enterprises. In aerospace, edge AI is already being used for in-flight data processing, reducing the reliance on ground systems. Applications include real-time navigation, on-orbit maneuvering of satellites, and predictive maintenance to lessen operational disruptions. For military applications, edge AI enables autonomous navigation for UAVs in GPS-denied environments and enhances situational awareness by processing sensor data locally on vehicles and drones. The global AI in aerospace and defense market was valued at $25.43 billion in 2024 and is projected to reach $65.43 billion by 2034. For resource-constrained aerospace systems, the choice between FPGAs and GPGPUs is critical. FPGAs offer low latency and power efficiency by processing data as it arrives, making them suitable for real-time inference at the edge. In contrast, GPUs excel at handling large batches of data for high-performance computation but can be less efficient for the small-batch inference common in edge applications. The open-source RISC-V instruction set architecture is also gaining traction for aerospace applications due to its customizability, which allows for workload-specific processors that are resource-efficient and fault-tolerant. NASA has selected a processor based on RISC-V for its next-generation space computing project. The integration of AI into Model-Based Systems Engineering (MBSE) is transforming aerospace design by automating the creation and validation of system models. AI can analyze system models to identify potential risks, suggest design trade-offs, and ensure requirements traceability. This AI-driven approach helps to manage the complexity of modern aerospace systems and accelerate development. Certifying AI and machine learning components for avionics under standards like DO-178C presents significant challenges. Current guidance treats AI-generated code the same as human-written code, requiring that every line be traceable to a specific requirement and be independently verified. Because of the non-deterministic nature of some AI, many organizations limit AI's role to an assistant for tasks like generating boilerplate code or suggesting test cases, with a human engineer remaining fully accountable. Full certification of AI for more critical functions may take years to become standardized.