It could be a matter of massively parallel computation (trillions of synapses functioning at once) plus efficiency of chemical-analog vs transistor. But, person

It could be a matter of massively parallel computation (trillions of synapses functioning at once) plus efficiency of chemical-analog vs transistor. But, personally I believe, there are a lot more algorithmic issues between what we are doing now and what the brain is doing. Even if it's just a matter of scale, we are at least a million fold away from the bandwidth and processing power of a single human brain vs exaflop supercomputers, especially in efficiency. Closing that gap, will require 20 years of Moore's law + bandwidth + efficiency + equivalent of 18 years grounded training (probably faster than 18y). We are not close to having our most powerful computers as flexible and general purpose intelligence as a single human. The 20 year limitation is based on processing, efficiency, and bandwidth. Hopefully the algorithms will catch up by then.