The idea of the “technological singularity” and the transhumanism movement is becoming more popular these days. Most notably, Ray Kurzweil proposes a “law of accelerating returns”, arguing that we should extend Moore’s Law (the number of transistors on a chip doubles every 2 years) to technology in general.
The problem with this is that Moore’s Law isn’t a scientific law, but folk wisdom. While it’s true that it’s held for approximately 40 years now, our ability to put more transistors on a chip could (and probably will) slow down. Either we’ll hit an upper limit due to physics, or society will simply lose interest, and move on to other things.
The latter will probably happen first. Airline flight speed is a perfect example of this. The “Spirit of St. Louis” flew from New York to France in 1927 in 33 hours. Today, a search on expedia.com reveals that the same flight still takes a little over 7 hours. If we applied a very liberal Moore’s Law to airline speed, saying that it doubled not every 2 years, but every 20 years, this flight would now take about 2 hours (a doubling every 2 years would put us at flight times of less than a second). Of course, we could very well do this – but it would be prohibitively expensive to implement at a commercial level. For now at least, people are happy taking 7 hours to cross the Atlantic.
Manned air flight records in general are even more telling. Up until 1962, there was a new record set almost every year since the first flight in 1903. Then there was one in 1965, then another in 1976. That record still stands. What happened? Why hasn’t manned flight speed been increasing exponentially? It’s simple – we as a society don’t care much about it anymore. Our focus has moved from aeronautics to fitting transistors on chips, and we’ve been doing that for roughly 40 years now. If interest wanes, then processor speed doubling is going to taper off very soon; perhaps as little as in 5 years.
A good case could be made for this happening sooner or later based on one simple fact: software needs to keep up in order to keep demand for better processors up. No one is going to buy a brand new 128 core processor when the software to utilize it wont be around for another 10 years (and when it does finally show up 10 years later, we’ll have 4096-core processors! to write software for!) Let’s looks at the release cycle of Microsoft Windows:
Windows 3.1 – April 1992
Windows 95 – August 1995
Windows 98 – June 1998
Windows 2k – February 2000
Windows XP – October 2001
Windows Vista – January 2007
Windows 7 – October 2009
WIndows 8 – May 2012 (preview version)
Except for a long gap between XP and Vista, what we see is a new major release every 3 years, not every 2. The point is that the speed of technology development waxes and wanes based on public interest. If software falls too far behind hardware, or if we simply lose interest in any improvement on both fronts, processor speed will stop doubling, just like advancements in aeronautics, construction, and many other fields have all but stopped. Moore’s Law, and by extension the law of accelerating returns, is simply a reflection of a current trend. And given the current state of AI and computing in general (not very close to perfection at all), I suspect it will be a long time before a technological singularity happens – almost certainly not in our lifetimes.