An argument against the Singularity

First off, WikiPedia is awesome. A brilliant idea.

WikiPedia “Technological Singularity“:

In future studies, a technological singularity is a predicted time at which technological progress accelerates beyond the ability of present-day humans to fully comprehend or predict. Such a singularity was first discussed in the 1950s, and vastly popularised in the 1980s by Vernor Vinge. People dispute exactly when the Singularity will occur. Predictions range from 2007 (Dan Clemmensen 1996) and 2012 (Terence McKenna 1996) to centuries from now to never. Futurists most commonly give the third decade of the 21st century.

More specifically, the technological singularity can refer to the advent of smarter-than-human intelligence (human or artificial), and the cascading technological progress (in nanotechnology and other areas) assumed to follow.

I like WikiPedia’s description of the Singularity as being a version of “The Rapture for Nerds”. Ray Kurzweil does some extrapolations in The Age of Spiritual Machines (which by the way is a very thought provoking book and I would recommend it. I discovered it during finals week one quarter and couldn’t put it down. I can lend it to anyone interested) which demonstrate that, in the past at least, certain areas of human development have experienced exponential growth from their inception to the present day. He reformulates this into “Kurzweil’s Law of Accelerating Returns” and uses it as a basis for the justification of the arrival of superhuman intelligence and the the Singularity sometime in the 21st century. This is obfustication. In fact, the Singularity really only depends on one exponential trend – Moore’s Law.

Here are some charts that give convincing evidence of humanity’s historically exponential progress in the domain of shuffling bits:

Given that Moore’s Law continues to hold until the point where it is possible for one machine to have the computational power of several human brains, the emergence of superhuman intelligence would appear inevitable. However, this supposition is dependent on the hypothesis that intelligence is a substrate-independent process. That is that it can occur outside the realm of human wetware. This does not seem to be an unreasonable assumption to make. In The Emperor’s New Mind, Roger Penrose suggests that human consciousness (quite different than intelligence, which we are discussing here) could depend on quantum processes inside neurons (thus preventing simulation on a classical computer). However, there is currently no evidence for this; Occam’s Razor applies. If we assume that intelligence is an algorithmic, substrate-independent process the problem of determining that algorithm remains. The evidence is that this is a very hard problem and will not be solved simply by throwing more transistors at it (Moore’s Law). Thus I take the position that while nothing, in principle, prevents the development of superhuman intelligence, even in the near future, it won’t happen any time soon because humans are too dumb to figure the algorithm out. Singularity is not inevitable.

However, this is not my main objection. The problem is that one of the key predictions of the Singularity theory is that the rate of increase in complexity and capability of human society is exponential. That just doesn’t mean “fast”, but rather that the rate of change of the rate of change is increasing linearly with time. Historically this has been true. Consider that the transition from horse carriages as a mode of transport in 1850 to the advent of the automobile around 1900 to the transition from that car to space travel in 1950. Progress in the domain of transportation here is superlinear. Kurzweil gives examples in his book in just about every domain you might think of. One essential domain that the Law of Accelerating Returns must support is energy usage, since any civilization uses power. Some clever Russian, Kardashev took this observation to it’s natural conclusion and invented a way to define the level of a civilization by it’s energy consumption. This is widely considered a reasonable idea.

The Kardashev Scale

Type I – A civilization that is able to harness all of the power available on a single planet, approximately 10^16 W. The actual figure is quite variable; Earth specifically has an available power of 1.74×10^17 W. Kardashev’s original definition was 4×10^12 W. (It was identified as a “Technological level close to the level presently attained on earth”, “presently” meaning 1964.)

Type II – A civilization that is able to harness all of the power available from a single star, approximately 10^26 W. Again, this figure is variable; the Sun outputs approximately 3.86×10^26 W. Kardashev’s original definition was 4×10^26 W.

Type III – A civilization that is able to harness all of the power available from a single galaxy, approximately 10^36 W. This figure is extremely variable, since galaxies vary widely in size. Kardashev’s original definition was 4×10^37 W.

I’m going somewhere with this I swear. Back to the Singularity theory. By this theory, if a civilization takes 40,000 years to become a Kardashev Type I civilization, then it should reach Type II status exponentially faster. Furthermore, the transition from Type II to Type III should be even quicker. Therefore, once intelligence arises someplace, any place, in the galaxy, it’s presence should be obvious. This observation is related to Fermi Paradox – from which one can potentially conclude that humans are the lone intelligence in the galaxy.

In fact, SETI has determined that there can be no Type II or III civilizations in our galaxy. There are also no Type I civilizations within 1,000 light years of earth in any direction. (Human’s energy production of 420 exajoules/year give us a Kardashev index of ~.65)

We can conclude:

If civilizations are prone to technological singularities we are alone in the galaxy.

And we must decide which is more likely:

1. There are physical/intellectual limits to technological progress that prevents the occurrence of singularity.

2. Life is exceedingly rare, with chances of existing in a single solar system being on the order of 1 in 100 billion. (Rare Earth Hypothesis).