Transcendent Man:

I don't think it's a secret that I'm not a Kurzweil fan. I once tried to pick my way through The Singularity is Near, but failed. Nevertheless, I'll definitely see the movie because I think it's interesting (did the same thing with What the Bleep, I'm not ashamed to say) even if I don't buy into many of Kurzweil's predictions. What I don't like, however, is to be a doubting Thomas for no reason. So here's my beef:

Number One:

The singularity isn't something I'm interested in discussing in detail, because it's essentially just the postulate from which all of Kurzweil's predictions come from. I don't necessarily object to the construct or the idea itself, I just don't believe that the preponderance of evidence supports this happening in the foreseeable future.

Number Two:

The law of accelerating returns is more like the law of diminishing returns. The favorite example of this is Moore's Law, which, I'm sorry to say, is visibly broken in its traditional form. We are butting up against the limits of the size of transistors we can make per unit area with current techniques (slightly over 30nm is the current cream of Silicon Valley), although graphene has been talked about quite a bit. Interestingly, the thing about graphene-scale transistors is that you're essentially getting down to scales where the same problems you have with quantum computers come into play, i.e. you're dealing with probabilities and quantum effects.

Some people think that processing cores will be the new Moore's law (i.e. that we'll have two cores in our computers, then four, then eight, etc). However, software to run on 8+ cores of hardware is prohibitively difficult from a programming perspective outside of rigid computational niches. I do actually agree with Kurzweil that things like optical and quantum computers may be the next big thing, but the idea that we'll have these supersmart computers to figure everything out for us isn't the whole story. If you've ever had experience with neural networks, you'll know that they are really just a set of heuristics for constructing systems of equations and programs to solve computationally difficult problems. They work for us because we have all this computer power available, and thus we can go a little further than we could've otherwise. But these methods are computationally dirty and WASTEFUL, and much like brute force cryptography is easily defeated by simple permutations (even with modern computers), so goes the notion of having computers do all our thinking.

Let me give you an example: a standard 128-bit key would take a computer checking 1 billion keys per second (let's pretend like that's what you could get out of a 1 GHz processor, just for the sake of argument) 1x10^22 years to try all the permutations. Ah, you say, my computer's MUCH faster than that. If you had a 10 GHz processor (which you don't, and again I'm simplifying the calculations here) you'd still be at 10^21 years. The age of the universe is only on the order of 10^10 years. The fastest supercomputer in the world (IBM Roadrunner) would be ~10^16 years in cracking the key (assuming 3 flops/Hz (AMD) in equivalency to previous calculation; again, there's some fuzzy math going on and oversimplification, but we're just playing with numbers here). Maybe this is all just a bad analogy, anyway, but as for having computers take over all our thinking, it seems like the permutations are just going to catch up with you somewhere along the line.

Number Three:
If the functionality of the brain becomes an open book in the next 20-30 years, it will be a happy accident. Believe me, I study the brain and I can tell you that nobody knows what's going on in any meaningful way. I, for one, think that if you want to create superintelligence, you should probably start by doing a better job of making people intelligent and focus on education instead of trying to create intelligent robots. Call me crazy. In fact, here's a prediction for you a la Kurzweil: as the gap between front-line science and what's taught in public schools grows wider and wider, pseudoscience will come to largely replace mysticism and other fringe beliefs as they struggle to maintain a place in the minds of people living in a modern technological world. Even the most intelligent are not immune to pseudoscience.

Number Four:
As to "living forever," I object strongly to rich people spending exorbitant sums of money on living forever when the child mortality rates are 18% in Angola, 15% in Sierra Leone, 15% in Afghanistan, 14% in Liberia, and almost 12% in Niger, to name but a few (these numbers come from The World Factbook). My philosophy is to make sure everybody gets a full share of a normal human lifespan before we go trying to live forever.

Anyway, it's always easy to be a naysayer but I'm willing to go conservatively along with Kurzweil far enough to say that something cool will probably happen within the next fifty years. What it'll be, I don't know. Go see Transcendent Man and maybe you'll get a glimpse, or maybe not. There'll be some cool shit to see in there anyhow.