Kurzweil essentially argues that the exponential progress seen in All Things Related to Brain Scanning Techniques and Subsequent Simulation of The Human Mind Based on Retrieved Data will eventually lead to the first "accurate" computer-based human mind (or human-based computer mind?) being created. This is a more likely scenario than the hypothetical event that the chaotic expansion process from compressed DNA code to the complexity of a "grown" human mind will be properly understood, at least prior to the genesis of a superhuman intelligence such as that which the technological singularity may result in. Either way, the computational requirements for simulation of the human mind must be met.
The IBM Roadrunner, which is currently the world's fastest computer (with grid computing not taken into account), already meets the lower bounds of the computational requirements for simulation of the functional equivalence of all human brain regions (10^14 cps). Neuron-model simulations of individual neurons, dendrites, synapses and spines (estimated at circa 10^19 cps) are currently not computationally feasible, but are most likely going to be possible on prospective supercomputers somewhat earlier than 2025, in conjunction with Moore's law.
I think grid computing approaches may make the technological singularity occur even earlier in the 2020's, although this would be under "optimal" conditions. In more realistic conditions, I think neo-luddite and anarcho-primitivist movements will grow stronger and oppose the prospect of "losing the legacy of humanity", thus resulting in delays. There is a possibility that they may grow so vocal that the fall of the industrial system will follow[1
], which may effectively make the singularity not happen at all. Such a scenario will happen at the expense of the lives of as many as 99% of the world's population due to the inherent population limits in hunter-gatherer societies[2
], although I consider this an unlikely scenario. There may also be various delays due to the difficulties in achieving a complete understanding of the functionalities of the human mind. To quote Emerson Pugh, "If the mind was simple to understand, we'd be too simple to understand it."
Vernor Vinge has stated, in the mid-1990's, that "[He'll] be surprised if [the technological singularity] occurs before 2005 or after 2030." Kurzweil's timeline defines the mid 2040's as when the singularity is most likely to happen. I.J. Good and Bill Joy agree on the intelligence explosion that a technological singularity involves, although Joy believes its aftermath will be dystopian.
Presupposing that a global nuclear war, a grey goo scenario or a supervolcano eruption will not take place, I expect progress in medical sciences to allow most of us to live long enough to live forever. Even in the event that detailed simulation of microtubules (miniscule components of the cytoskeleton) will be required to achieve the singularity, which may consequently force us to wait (from the point of view of an individual, due to the inherent co-operative nature of exponential technological progress) until the 2060's or 2070's until the required computational capacity becomes available, it is feasible that a majority of the people who live today will still be around. Now, if we need to achieve quantum-level simulations of the human mind, with an estimated peak capacity of 10^42 cps, it would, as extremely unlikely as such a massive requirement for computational capacity is, be problematic.