“@shivon There is a physics argument that synapse activations take 1 to 2 orders of magnitude less energy than silicon transistors. That, of course, does not explain why a 10MW GPU cluster still cannot write a better novel than ~10W of brain power. My guess is that silicon…”