[ExI] Serious topic
anders at aleph.se
Mon Feb 28 23:40:54 UTC 2011
Eugen Leitl wrote:
> The only way to give Moore second wind is to leap into 3d, and
> I'm very leery expecting just-in-time new technologies.
The real Moore's law is flops per dollar rather than transistors,
anyway. Right now it runs on increasing cores, it seems.
I think *Sahal's law* is strong enough: given the still remaining
enormous demand for processors of all sizes, we will see continued
exponential improvement for at least two more decades.
>> Just like Bass technology diffusion curves they fit data very well in
>> retrospect. But when used to extrapolate the future they seem too
>> unstable: the predicted peaks or sigmoids jump all over the place due to
>> noise in the data.
> I see little magic in graphs, but in fundamental processes giving
> rise to the graphs.
Exactly. But that rarely gives insight enough to make the quantitative
predictions we care about. The fact that there is a finite amount of oil
(modulo slow abiogenesis) doesn't tell us whether we should be heading
for the hills, investing in solar power or form a bike gang. That Bass
curve dynamics is well understood will not help you predict how many
gizmos you will eventually sell.
> Markets are chronically poor long-term predictors.
Maybe. But they seem to be extremely quick at changing when there is a
constraint. Remember what happened to the energy use per capita vs gdp
per capita post WWII?
A very strong linear relation from 1949 to 1970, followed by what has
essentially been a horizontal curve (growing GDP with no more energy
use). The reason was the oil crisis. It induced a surprisingly quick
shift of a long-term, heavily invested infrastructure.
Future of Humanity Institute
James Martin 21st Century School
More information about the extropy-chat