根据目前的神经科学，相比计算机的高功耗，大脑能在低功耗上完成许多壮举主要来源于三个方面：
vast connectivity, structural and functional organizational hierarchy, and time-dependent neuronal and synaptic functionality。
1、广泛的连通性
2、结构上和功能上有组织的分层系统
3、时间依赖性的神经和突触信息传递功能
虽然现在的DLNs是仿照神经网络设计，但是和真正的大脑仍然存在不同之处：
(1) the segregation of computations (the processing unit) and storage (the memory unit) in
computers contrasts with the co-located computing (neurons) and storage (synapses) mechanisms found in the brain;
(2) the massive threedimensional connectivity in the brain is currently beyond the reach
of silicon technology, which is limited by two-dimensional connections and finite number of interconnecting metal layers and routeing
protocols;
(3) transistors are largely used as switches to construct
deterministic Boolean (digital) circuits, in contrast to the spike-based
event-driven computations in the brain that are inherently stochastic
1、硅基计算系统的存储单元和运算单元分开，大脑的存算一体
2、大脑的3维立体复杂连接网络受限于硅基科技的2维连接和有限的路由协议、互连神经层级。
3、被设计用于人工智能的电子电路是固定不变的，而大脑中事件驱动性的基于脉冲的计算系统是随机改变的。
虽然目前硅基计算系统平台（如图像识别）已经成为人工智能革命的可能因素，但是其高功耗的瓶颈阻止了它的进一步普及。