Sunday, January 28, 2007
Post on Moore's law as made to slashdot.
if I'm posting at slashdot, I'm probably not getting my work done.
In response to
Re:Dupe reply - why is moore's law a law?(Score:2)by arbitraryaardvark (845916) <gtbear.gmail@com> on Sunday January 28, @05:02PM (#17792484)
"Why it's Moore's Law a law? It just sounds like a theory to me, it just has been surprisingly accurate to date, that's all."
Theories that remain suprisingly accurate over time tend to be known as laws. Unlike, say, axioms, where one counterexample could break a paradigm, a law only has to work often enough to be useful. If a prediction works 95% of the time,and fails to account for 5% of the data, we can still call that a law. Feel free to call it Moore's pretty damn good conjecture. It's not intended to be rigorous,and we don't need to claim that that it will work for the next 10,000 years. It's enough to understand the general point that the cost of an information processing system is cut in half every two years or so by developing technology, and that can only have profound changes on culture and economy.
It's useful to be familar with a couple of additional concepts: 1) austrian economics, which shows how markets function to drive technological change,and how technological change functions to drive markets. 2) the singularity. aka "the rapture for nerds", the singularity
is the idea that the rate of technological change is speeding up, driving innovation in ever shorter cycles, in a hyperbolic curve (y=x squared), so that at some point, probably in this century, the rate of change will be going basicly straight up,and that on the other side of the singularity, things look weird.
So there are at least three options:
1) Moore's law is an overstatement in the long term. At some point physical limitations set in, the low hanging fruit has already been picked,and a new plateau is reached where the cost of information systems is low compared to today, but has leveled out and is no longer decreasing.
2) Moore's law will continue to be suprisingly accurate for many years to come.
The cost of information systems will keep decreasing by about half every two years,and that will continue to drive economic transformation and social change.
3) Moore's law is descriptive at the elbow of the curve, where we live now, but as change builds on change Moore's law will be found to be wildly conservative,and the cost of a given information system will drop by half in shorter and shorter cycles, until information system costs approach zero, with consequences that include AI, space travel, life extension,
gene hacking, and stuff we can barely imagine now.
As formally stated in terms of doubling of transisters on a chip, or in terms of the cost of a transister, per period of time, moore's law only applies to the time since the invention of the transister and some unknown point in the future at which it no longer applies, perhaps because we use something else besides transisters. It remains useful in describing the period
from about 1950 (or 1970) through 2007 up to at least until either limits are reached or the pre-singularity effects kick in and shorten the doubling time. I expect measurable pre-singularity effects by 2012. Some would argue Moore's law is itself an example of noticable pre-singularity effects.
4) two cups of Moore, 1/2 cup of salad dressing = moore slaw
if I'm posting at slashdot, I'm probably not getting my work done.
In response to
Re:Dupe reply - why is moore's law a law?
(Score:2)by arbitraryaardvark (845916) <gtbear.gmail@com> on Sunday January 28, @05:02PM (#17792484)
(http://vark.blogspot.com/ | Last Journal: Thursday February 17, @07:49PM)
Theories that remain suprisingly accurate over time tend to be known as laws. Unlike, say, axioms, where one counterexample could break a paradigm, a law only has to work often enough to be useful. If a prediction works 95% of the time,and fails to account for 5% of the data, we can still call that a law. Feel free to call it Moore's pretty damn good conjecture. It's not intended to be rigorous,and we don't need to claim that that it will work for the next 10,000 years. It's enough to understand the general point that the cost of an information processing system is cut in half every two years or so by developing technology, and that can only have profound changes on culture and economy.
It's useful to be familar with a couple of additional concepts: 1) austrian economics, which shows how markets function to drive technological change,and how technological change functions to drive markets. 2) the singularity. aka "the rapture for nerds", the singularity
is the idea that the rate of technological change is speeding up, driving innovation in ever shorter cycles, in a hyperbolic curve (y=x squared), so that at some point, probably in this century, the rate of change will be going basicly straight up,and that on the other side of the singularity, things look weird.
So there are at least three options:
1) Moore's law is an overstatement in the long term. At some point physical limitations set in, the low hanging fruit has already been picked,and a new plateau is reached where the cost of information systems is low compared to today, but has leveled out and is no longer decreasing.
2) Moore's law will continue to be suprisingly accurate for many years to come.
The cost of information systems will keep decreasing by about half every two years,and that will continue to drive economic transformation and social change.
3) Moore's law is descriptive at the elbow of the curve, where we live now, but as change builds on change Moore's law will be found to be wildly conservative,and the cost of a given information system will drop by half in shorter and shorter cycles, until information system costs approach zero, with consequences that include AI, space travel, life extension,
gene hacking, and stuff we can barely imagine now.
As formally stated in terms of doubling of transisters on a chip, or in terms of the cost of a transister, per period of time, moore's law only applies to the time since the invention of the transister and some unknown point in the future at which it no longer applies, perhaps because we use something else besides transisters. It remains useful in describing the period
from about 1950 (or 1970) through 2007 up to at least until either limits are reached or the pre-singularity effects kick in and shorten the doubling time. I expect measurable pre-singularity effects by 2012. Some would argue Moore's law is itself an example of noticable pre-singularity effects.
4) two cups of Moore, 1/2 cup of salad dressing = moore slaw
Comments:
Post a Comment
