Microsoft Word Kurzweil, Ray The Singularity Is Near doc

Software Stability.
Lanier calls software inherently "unwieldy" and "brittle" and has described at great length a 
variety of frustrations that he has encountered in using it. He writes that "getting computers to perform specific tasks 
of significant complexity in a reliable but modifiable way, without crashes or security breaches, is essentially 
impossible."
3
It is not my intention to defend all software, but it's not true that complex software is necessarily brittle 
and prone to catastrophic breakdown. Many examples of complex mission-critical software operate with very few, if 
any, breakdowns: for example, the sophisticated software programs that control an increasing percentage of airplane 
landings, monitor patients in critical-care facilities, guide intelligent weapons, control the investment of billions of 
dollars in automated pattern recognition-based hedge funds, and serve many other functions.
4
I am not aware of any 
airplane crashes that have been caused by failures of automated landing software; the same, however, cannot be said 
for human reliability. 
Software Responsiveness.
Lanier complains that "computer user interfaces tend to respond more slowly to user 
interface events, such as a key press, than they did fifteen years earlier....What's gone wrong?"
5
I would invite Lanier 
to attempt using an old computer today. Even if we put aside the difficulty of setting one up (which is a different 
issue), he has forgotten just how unresponsive, unwieldy, and limited they were. Try getting some real work done to 
today's standards with twenty-year-old personal-computer software. It's simply not true to say that the old software 
was better in any qualitative or quantitative sense. 
Although it's always possible to find poor-quality design, response delays, when they occur, are generally the 
result of new features and functions. If users were willing to freeze the functionality of their software, the ongoing 
exponential growth of computing speed and memory would quickly eliminate software-response delays. But the 
market demands ever-expanded capability. Twenty years ago there were no search engines or any other integration 
with the World Wide Web (indeed, there was no Web), only primitive language, formatting, and multimedia tools, and 
so on. So functionality always stays on the edge of what's feasible. 
This romancing of software from years or decades ago is comparable to people's idyllic view of life hundreds of 
years ago, when people were "unencumbered" by the frustrations of working with machines. Life was unfettered, 
perhaps, but it was also short, labor-intensive, poverty filled, and disease and disaster prone.