[ExI] [mta] Re: Bitcion Moore's Law?

[Eugen Leitl]

On Wed, Jul 17, 2013 at 12:07:08PM +0200, Anders Sandberg wrote:
> On 2013-07-17 10:54, Eugen Leitl wrote:
> >On Mon, Jul 15, 2013 at 02:14:07PM -0600, Brent Allsop wrote:
> >
> >>Moore's law is simply a small and specific example of the more general
> >Moore's law is over. Financial scaling ceiling is here, physical
> >is a couple of nodes away.
> 
> I think you both are wrong, at least in general terms. What is

Moore's financial scaling limits have been reached. That's no
longer a prediction about the future, it is now a statement
about the past, just as Peak Oil is. 

> driving the improvement of many technologies is Wrightean learning
> and massive expansion of uses as the unit price goes down:

In Moore, the unit price is no longer goes down, it's now
flat. The financial incentive for low-margin products like
mobiles and exascale is now no longer there.

The physical limits are not far behind, and rediced incentive
implies the linear semilog is no longer a straight line.

> http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0052669

Yeah, I'm aware of that paper.

> If there is no interest in a technology, it is not going to go
> Moore/accelerating returns (and surprisingly many techs like

Linear semi-log plot technologies are really rare, and limited
in their temporal scope. Why some think that they're the rule,
and go on forever is probably due to a yet another failure of
exponential returns: education.

> capability based security are stuck here). Physical limitations are
> surprisingly weak influences on price, since price is more about

In low-margin markets, physical limitations have everything
do with prices. A mainframe is not a low-end mobile phone.

> what people pay rather than anything objective - and often later
> versions of the tech are fundamentally something different from the

Moore applies to a given technology. There are no alternative
technologies available to neatly take over no more than when
system clocks stopped doubling a decade ago:

http://www.extremetech.com/wp-content/uploads/2012/02/CPU-Scaling.jpg

The dark blue line remained flat. So will the green line.
Further progress can no longer depend on four times as
many widgets with halfed widget feature size. The doubling
times are no longer constant, they're increasing.

> earlier versions of the "same" tech (candles vs. gaslights vs.
> incandescent lightbulbs vs. LEDs - same goal, but very different
> ways of achieving it, and with very different ultimate physical
> constraints)

Moore is about semiconductors. When the curve stops being
a linear semi-log, the law is dead. That eventually some
other substrate and technology comes along that has a different
scaling is immaterial to the original issue: that Moore is dead.
 
> The scaling ceiling in economy is likely set by how well and large
> we can coordinate capital and groups of people. This is something

I have further bad news for you: not only is Moore dead, so
is the economy. GDP is largely a fabrication, but even that
fabrication follows availability of cheap energy. There was
never a decoupling. Negawatts don't really exist.

As this planet did not consider that availability of cheap
energy is a high priority in the last 40 years, the next 30-40
years will have flat to contracting growth (shrinkage).

> that seems to improve over time, and I suspect it does in a
> multiplicative fashion.

Infrastructure transitions always take 30-40 years. 
I'm going to argue that the capital for that is no longer there,
and the skilled manpower is also no longer there (at least 
in Europe and North America).