Pretty cool (extremely verbose response)

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Pretty cool (extremely verbose response)

Giles Bowkett
Personally I have mixed feelings about this.

There's actually a search engine startup in Albuquerque which is
actually that one tiny bit better than Google. I kind of expect Google
to buy them up sooner or later.

http://www.gigablast.com/

The thing is, their web site says that they hope to outdo Google.
That's ridiculous. Google's real asset isn't its search; it's the
Google filesystem. GFS holds nearly the entire Web in RAM. This
enables Google to be incredibly fast. Even without the brand
recognition, you can't displace Google without an equivalent or better
filesystem. Because of this I think Google might actually be a lot
more similar to Microsoft, in the long term, than you would expect.
Their principle asset is essentially an operating system.

They're already building a data-center-in-a-box. It's already possible
to pick up a Linux box for $200 which is the size of a keychain. In
ten years, it might be possible to pick up a **data center** for $200
(or the equivalent) which is the size of a keychain. At that point,
data centers will be a commodity, and Google will be in the business
of controlling access to the operating system all these commodities
require.

That's Microsoft's business model. The difference is that the initial
technology Google started with was genuinely the best in the field,
while you can't say that about Microsoft, but its ultimate destiny
could well be very similar.

Google's positioned to control the world, as long as the world keeps
using information the way it does. However, advances in robotics make
it highly likely that the amount of information exchanged over the
Internet in ten to twenty years is going to be an extremely large
number of orders of magnitude larger than the current amount of
information exchanged over the Internet. It's not just that human use
will increase, or scale to multimedia; the minute you get a safe
self-driving car, accidents will disappear, insurance companies will
scale down to the size of gnats, the economic efficiency of all
industries which use shipping in any manner will increase
dramatically, and the amount of information exchanged over the
Internet will mushroom to a volume unimaginable today.

Rodney Brooks is already shipping prototype self-driving robot
vehicles to Iraq and Afghanistan for the US military. In Japan,
prototype robot security dogs with built-in webcams are already
patrolling a few very expensive corporate offices. It's only going to
take time for advances in robotics to create huge demand for commodity
data centers.

Many of the most effective artificial intelligence techniques are
statistical techniques which require extremely large volumes of data.
Machine translation is a good example; Google has an excellent working
prototype that can translate from Arabic to English without errors,
performing much, much better than the industry standard:

http://www.csmonitor.com/2005/0602/p13s02-stct.html

There are two reasons for this superior performance: superior
engineers, and the Google filesystem, which allows those superior
engineers to perform statistical analyses on much, much larger volumes
of data than ever before. Fast forward about ten years. If you have a
commodity data center the size of a keychain capable of storing
Google-equivalent data ranges, bam, you've got a universal translator
which fits in a keychain.

Of course it'd make more sense just to use a WiFi box that gives you
the same result without waiting a decade for the technology to
miniaturize, but the point here is the inevitable demand for commodity
data centers. It would be extremely useful for self-driving vehicles
to be able to process equivalent data ranges with equivalent or
superior results. Little baby robot trucks are already following
soldiers and humvees all over the deserts of Iraq. Those little baby
robot trucks might not actually have any business being there, any
more than the soldiers or the humvees, but that's another topic
entirely, and you can hardly blame a robot for being in the wrong
place at the wrong time. The point is that they exist, and if they can
already follow humvees and soldiers across rocky terrain without
crashing into the humvees or running over the soldiers, putting them
on the highway isn't an if, it's just a when. And when that when hits,
these things are going to want to process and exchange a **lot** of
data. The first robot truck to traverse the American highways will
definitely have a human driver at the wheel just in case anything goes
wrong, but it could also very well be carrying a Google data center in
a box as its payload. The more robot trucks, the stronger the impetus
for those data centers to miniaturize, and the more they miniaturize,
the more useful they become. But when this happens, those data centers
become commodities, and Google becomes the next Microsoft.

(That being said, they'll probably do a lot of cool stuff before that
happens. I mean they've got time. They're not under the gun or
anything.)

On 11/20/05, Douglas Roberts <doug at parrot-farm.net> wrote:

> Bingo.
>
> On 11/20/05, Roger Critchlow <rec at elf.org> wrote:
> > World domination for less than Microsoft spends on legal settlements,
> > sounds good to me.
> >
> >
>
> --
> Doug Roberts
> 505-455-7333 - Office
> 505-670-8195 - Cell
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at Mission Cafe
> Wed Lecture schedule, archives, unsubscribe, maps, etc. at
> http://www.friam.org
>
>


--
Giles Bowkett = Giles Goat Boy
http://www.gilesgoatboy.org/