Researchers and tech firms are turning to decentralized AI training to curb the rising energy demands of large models, distributing workloads across existing, underused computing resources instead of ...