the first problem is that there are so many conceptions of the singularity.. instead of dealing with any particular one here I'll try to generalize
it seems that the most common footing for singularity predictions is moore's law, whose exponential curve gives us the misnomer "singularity"
personally, I don't like infinities... my own natural intuition tells me there are no infinities in the universe, so when I see this curve of progress heading toward heaven I think there's a problem with it
from the simplest point of view, it makes sense that technology should be exponential, because we can use each generation's technology to make the next generation's tech, which should keep everything going faster (*points to kurzweil's the age of spiritual machines*)
but I think there are two other possibilities (ok, at least two more but here are two I have slight reason to believe)
the first is that the curve may become linear, because there's always a delay between the introduction of a technology and it's use, which may eventually lead to a saturation.. similarly, eventually so many articles will be published in journals each year that it'll be a full time job to fail to keep up with the newest techniques and results
of course, these same new technologies can be applied to the distribution of knowledge, so perhaps this problem will fix itself
my other reservation is that all these sigmoid curves adding up to an exponential curve doesn't strike me as correct... if we can say that the curve will stay exponential because we assume any problems will be fixed along the way, I think it's just as valid to say that we're really approaching the inflection point on a sigmoid curve, because self-similarity seems to be a popular trait in the universe: why shouldn't sigmoids add up to a sigmoid instead of an exponential?
I've kindly asked mathcad to regress some year vs. mips data to a logistic curve, but alas the results are so sensitive to the guess values that I can't get anything consistent out... maybe I need to use different data
I must conclude by mentioning that neither of these concerns makes the singularity impossible necessarily, they just delay it (which I don't like because I want to be alive for it)