To my understanding, pyramidal cells are already partially emulated by certain structures of weights. It's not the same as dedicated multi-layer mechanisms outs

To my understanding, pyramidal cells are already partially emulated by certain structures of weights. It's not the same as dedicated multi-layer mechanisms outside the local flow, but that option seems difficult to implement efficiently with current compute technology. Maybe more VRAM to hold the added inter-layer features would allow it? NVidia seems reluctant or unable to do that on the consumer side. I haven't had time to thoroughly read up on why. All of these seem plausible: * Offsetting rising per-unit costs from other components * Intentional market segmentation * Export restrictions