What's interesting about the above thought is it supposes that what have the technology now, we just lack the technique. You might disagree, but what you wrote paved over the essence of the parent.
I see what you mean. I'd read through the comments at the linked page, and I'd forgotten that the linking comment here on HN didn't include those debates.
Especially, whether you can pack in a human's worth of knowledge into a 2005 HDD (of course you can if you have huge numbers of them, but seems against the spirit of the idea).
And, the estimates of human neural complexity. It's possible that the human brain isn't very efficient at intelligence alone - there's many survival functions and the historical path evolution happened to take. (as harder evidence, I also read of a guy with almost no brain, just a few millimeters coating the skull, who had normal college IQ...)
But I was really adding my own tangent, a thought I had long ago: assuming intelligence can be simulated by an algorithm, it can be done now, because of turing machine equivalence (provided we have sufficient memory). If it used too much memory, it seems likely we could still run a cut-down version, or some aspect of it. [BTW: Arguably, if a mathematical model of consciousness exists, with variations for personality, knowledge, present state of mind etc, all of them are virtually conscious and feeling at all times - just frozen in stasis at the moment].
Historically, one barrier to new algorithms is that practical people just don't entertain algorithms that are computationally infeasibile on present hardware. But when hardware is fast enough, they are more willing to. Perhaps the ability to experiment and iterate quickly is also helpful.
