Training-Inference Bifurcation
Training stays centralized (Leontief); inference decentralizes (substitutable).
Impact Score
Score Reasoning
- Importance
- Central to Paper 6 (AI Transition): training-inference split is the structural mechanism for AI decentralization. Connects CES complementarity to AI industry structure.
- Novelty
- New structural explanation of AI centralization/decentralization via CES rho: training is Leontief, inference is substitutable. The Baumol bottleneck connection is a novel cross-domain link.
- Quality
- Comprehensive article covering training complementarity, inference substitutability, Baumol bottleneck, and self-undermining dynamic. Good cross-links.