Saturday, September 8, 2012
Sliding Toward The Singularity
Intel's co-founder Gordon E. Moore noted in 1965 that the components in integrated circuits doubled every two years. He proposed that this would continue and it has. This doubling of transistors in CPUs has lead to an increase in processing power that doubles about every 18 months. This is why today your cell phone has more processing power than the best 1965 mainframe. So what does that mean for the future? In theory it means the processing power of your lap top could fit in your wrist watch just a few years from now. In practice its not that straight forward.
Moore's Law is confined by physical limits and is already showing signs of faltering. At this time chips are being made that have widths of 20 atoms. Obviously you can't possibly go below one atom with a standard silicon chip, but realistically you can't even go below five since the heat created would be enough to melt the chip. There's also the problem of leakage in the wiring between transistors. Due to quantum probability the electron sometimes falls outside such a thin wire. For the last few years the chip makers have been using multiple cores to make up for the inability to increase processing speeds at the previous rate. This is not as good as having a single core that is faster. One reason for this is the single threaded nature of most programming. So what's next?
Intel is trying to keep the rate of progress going by using 3D chips, AKA Tri-Gate chips. These may yield some increase for a time, but in order to keep the rate of progress on track for the long run we need a completely different kind of computing. Several strategies are being explored with limited success. Some of them involve using the quantum phenomena of spin others use biological things like DNA. There's also some more immediately attainable technologies involving the use of graphene. The idea there being that by switching from silicon to materials like graphene, which carries electricity faster than silicon, far less heat is created and thus power consumed.
Power consumption is a major issue, in fact the physical limits imposed by the availability of electric power and the copper wire/fiber optic cable the net is linked together with are also showing signs of being unable to keep up with the current rate of technological proliferation. In particular the demand for multi-media is increasing so fast that it will require equally significant increases in the through put of communication networks and decreases in power consumption by servers.
Too Much Magic
It has been said that technology sufficiently advanced is indistinguishable from magic. Is it any wonder then that many who have witnessed this astonishing rate of technological progress believe that this rate of progress will magically continue and further solve all our problems. James Howard Kunstler’s book Too Much Magic talks at length about the unfounded optimism that people have regarding the future. When confronted with some of the real and very dire problems facing society there seems to be a common belief that the magic wand of Technology will save us. Often ideas like hydrogen cars and vertical farms are touted as solutions to the problem of resource scarcity without any thought to how these things will actually work or any regard to the arithmetic of the closed system in which we exist. Likewise with the future of computers and related technology there is an almost magical belief that the conjurers, AKA scientists and technocrats, will come up with a solution.
Many of these physical limits will likely prove impervious to technological advancement. There is only so much energy available on this planet and you can only make a wire, chip, transistor, etc so small. At some point, and I think we are there, that exponential hockey stick will curve into a plateau of diminishing returns. I personally don't think that's such a bad thing.