One of Kurzweil's claims is that there have been five computational paradigms over the last 110 years: mechanical switches, relays, vacuum tubes, transistors and integrated circuits.  Each of these develops until it reaches a plateau which then spurs the next paradigm.  There are some problems with this:

1.  The definition of 'paradigm' is unclear.  For instance, there isn't a large difference between mechanical switches and relays yet Kurzweil separates them into two separate paradigms.  Also, transistors and ICs are based on the same physical phenomenon which means they could be considered the same paradigm.  Conversely, each generation of ICs could also be considered a unique paradigm.  You could have 2 paradigms (information stored by mechanical position or electronic charge) or 20 paradigms (... 250nm, 180 nm130 nm, 90 nm, 65 nm, etc...).  Too few would mean that the concept of paradigms can't be used to extrapolate, too many and it loses its significance.  Kurzweil arbitrarily chooses five paradigms is a nice Goldilocks number. 

2.  All technologies have paradigms, even the ones that have reached a plateau like aviation.  Aviation has had at least 3: propeller, rocket and turbine.  Simply saying the there have been paradigms doesn't mean that there will be a new one to come along.

3.  The earlier transitions occurred when the computer industry was very small and so it was much cheaper to retool to the next technology.  I don't have numbers but the amount of money necessary to move from relays to vacuum tubes must have been miniscule given the fact that the industry barely existed at the time.  Today, the semiconductor industry is a >1 trillion dollar global industry which has enormous inertia that must be overcome to retool for a new 'paradigm'.