Most of what is presented as opposition between a gating network and basic prediction methods in venture capital is noise.
They are complementary techniques where a gating network can refine the results of existing methods, not contradict them.
Control lies in pragmatic integration, not in artificial distinctions of competition.
For example, an LLM can generate text embeddings that are then weighted by a gating network for an improved prediction of startups.