> To be fair, today we have what we believe to be (much) more accurate models of how much computation a human mind is capable of
How much, yes, how, not so much and definitely not at the powerbudget the brain has.
> and how long it will take to build computing machines operating at that scale.
We don't actually know that. There have been some WAGs but so far those appeared to be totally off based on the developments since.
> 36 years ago the argument that AGI was coming soon could be made in tandem with the argument that we'd make some fundamental advance that allowed computers to express intelligence with less computational capacity than humans (by orders of magnitude).
Yes, that was a crucial mistake and it led directly to the AI winter.
> Today we can make an argument that we'll achieve it (at least initially) by leveraging computational capacity on par with or orders of magnitude greater than a human mind.
Chances are that we're missing a very important piece of the puzzle for which there is no known solution even in theory. The problem is that there are many candidates for that important piece none of which have currently proposed workable solutions no matter what the computational budget or the accepted slowdown (they are equivalent).
So I think some caution when throwing around projections numbering 'just a couple of years' is warranted, after all it's 'merely a matter of programming' but in this case we don't have a working model that we understand.