My point, of course, was that an animal contains within itself the blueprints not only for all the “hardware” components that make up its own body, but also the instructions for self-assembly (development), as well as “software” (instinctual behavior) that will allow it to produce other individuals (obtaining food for growth, mating behavior). Just because it is highly automatic doesn’t mean it is simple. The reason the iPhone is hard to make is because it’s not complex enough; it’s incapable of self-assembly.
Samuel Butler covered this objection: machines need humans as part of their reproductive cycle just like certain plants use insects for pollination.
Build an iPhone out of jelly and string, and show it to the class on Monday.
I think some of the high estimates are of how much conventional computer power it would take to simulate human brain — not the same thing.
As a quick back-of-envelope, consider that the brain has about 200 trillion synapses, each triggered an average of 15 times per second, say. That’s 3 quadrillion multiplications per second. But precision is low, redundancy high, reliability doubtful, and most of the power will sit unused for any given task. Round it to 10^12 FLOPs as a very crude guesstimate, though the differences may be too great for comparison to be meaningful. (OTOH, there are over 10^18 tubulin-dimers in the brain, so that guesstimate would be low if, as Penrose and others speculate [pdf], the microtubules do significant processing.)
To make this an apples-to-apples comparison, the computer would have to view each digit, run it through an OCR subroutine, access an associative memory for the 10x10 multiplication table, and so on. And do all this while most of its compute power is spent on maintaining posture and vigilance. (Some of this is bypassed in idiot savants who multiply orders of magnitude faster than ordinary humans.)
It will not be like The Onion until and unless they start publishing articles like this: