[extropy-chat] and the nano/holo fun goes on...

[Emlyn]

On Sun, 15 Aug 2004 21:47:01 -0400, Dan Clemmensen <dgc at cox.net> wrote:
> Don't worry, there are a lot of drivers for investment in performance
> increases
> in semiconductors and storage, and the phenomenon you worry about is already
> pervasive.
> 
> You can apply more computational power to a problem in either or two ways:
> you can get faster processors, or more processors. We are getting
> increasingly better
> at distributed computing. Yes it's hard,a nd yes, it works better for
> some problems
> than for others. But our canonical worked example for intelligence, the
> biological brain,
> appears to be a massively distributed implementation.

We need to be looking at new solutions to the software problem too
(ie: the problem that software technology is not increasing in-line
with hardware, because it's still hand-coded). As we get gruntier
hardware, hopefully we'll begin to see stupidly wasteful software
generators, based on things like genetic algorithms. So we can start
moving further down the track of all spec, no manual implementation
software, which will change things dramatically.

> 
> For disk drives, we see exactly the "problem" you mention. Drives have
> quit doubling
> in capacity, and instead are now dropping in price, specifically because
> a nominal 1-drive
> system has "enough" capacity.  You should look at
>   http://www.pricewatch.com
> occasionally to get a feel for this trend. Disk capacity improvement
> have consistently
> outpaced processors performance improvements for at least the last five
> years. The current
> "sweet spot" for cost/GB is at about 250GB for $.50/GB, at retail.  Add
> $500 for a cheap
> computer to house four drives: you get 1TB for $1000.

I figured out you could do a TB for $1000 a couple of weeks ago... it
blows my mind.

By the way, I still use up all my drive space (eg: 120gb drives),
mostly due to VMWare. I use virtual machines with abandon for
developing and testing distributed systems. Powerful stuff!

> 
> The only way I know to use this much disk at home is to store DVDs.
> Storing audio
> CDs hardly counts: my entire CD collection fits on <<300GB with lossless
> compression.
> You can store about 5 audio CDs with lossless compression (flac) in one
> GB, for a storage
> cost of $.20/CD. Storage cost for a DVD would be about $4.00, which is a
> small fraction
> the purchase price of a DVD. In my case I have a legal, physical CD for
> each of the CD images
> on disk, but you can see why the RIAA and MPAA are panicking.

They are panicking because they're struggling to wring the last drops
out of last decade's business model, but they don't have a good
alternative for this decade. I can tell them what it is, for free...
run sites like the russian mp3 download sites, selling songs for 10c
each, or albums for $1, and make it fast and easy to download. Problem
is they've taken too long to set it up, and now the pirates have
beaten them to the punch. Poor old recording industry, boo hoo.

Or, hey I know, why don't they just piss off and leave artists and
audience to interact directly online?

> 
> If the semiconductor manufacturers move off the performance curve, they
> will simply move
> to the price curve instead. We will see retail computers at $100. 

Yeah! Bring it on!

> With
> the semiconductor content
> dropping in price each year. due to intense competition on margins. But
> this means that
> the cost of building a cluster will drop dramatically, so the
> singularity continues to advance
> at its inexorable pace.
> 

Don't worry Kevin. We'll keep finding ways to use all that mighty
hardware, I promise.

-- 
Emlyn

http://emlynoregan.com   * blogs * music * software *