The only way to get rid of centralized choke points is to actually go decentralized. At Q Blocks, we're working on making this solution a reality for a lot of the ML devs constrained by the computing costs on cloud.