Apple’s research project splits AI inference workloads across connected Macs using Apple Silicon, and thereby created a shared pool of graphics units capacity