Follow the incentive: who benefits from the idea that a gating network is fundamentally opposed to basic methods in venture capital?
This formulation creates a conflict of interest that does not exist, because a gating network cannot operate without these basic methods (GNN, embeddings, LLM).
The network merely orchestrates and refines signals; if the data or basic models are biased, the network amplifies these flaws.
For example, if initial models systematically underestimate a startup's liquidity risk, gating will not correct this fundamental gap but propagate it.
Real progress requires improving the foundations, not a false dichotomy.