A research team is deciding how to allocate resources for the next six months to improve their language model. They have two options:

• Option A: Focus on complex architectural modifications and specialized training techniques for the current, medium-sized model.
• Option B: Use all available resources to significantly increase the model's parameter count and the volume of its training data, keeping the fundamental architecture the same.

A senior researcher argues for Option A, stating, "Simply making the model bigger is a crude approach. We won't see any fundamentally new behaviors, just slightly better performance on tasks it can already do."

Based on observations from the development of very large models, construct a counter-argument to the senior researcher, justifying why Option B is a compelling strategy.