Microsoft Patent Targets Idle GPUs for AI Training with Rewards
Microsoft has published a patent detailing a system that harnesses idle GPU power from user edge devices to perform micro-training tasks for AI models. In exchange for contributing their computational resources, users can earn points redeemable for free software, model access, or cash. This novel approach addresses the persistent industry-wide shortage of AI hardware and soaring infrastructure costs by creating a decentralized cloud computing network. By acting as an intermediary, Microsoft could help cash-strapped companies and researchers train AI models using distributed consumer hardware. The patented system continuously monitors hardware metrics such as CPU and GPU utilization, memory usage, temperature, and power consumption to run tasks opportunistically without impacting user experience. However, the system is currently only at the patent stage, and questions regarding user privacy and hardware wear remain unaddressed.
## BACKGROUND
Distributed computing involves pooling resources across multiple networked devices to handle large-scale computational tasks that would otherwise require expensive, centralized data centers. Edge computing further refines this by shifting data processing closer to where the data is generated, reducing latency and utilizing local hardware.