[ExI] The Actual Visionary of the Future
eugen at leitl.org
Tue Oct 29 11:29:28 UTC 2013
On Mon, Oct 28, 2013 at 01:51:22PM -0600, Kelly Anderson wrote:
> I think his data is well researched. Whether all of the curves extend into
> the future, and just how far they will extend is guess work.
No, no, no. If you're formulating a theory, you have to define
its scope of applicability, and margins beyond which the
theory is falsified.
If it's not a theory, then why are we wasting time on
> > > believe everything is exponential. That being said, lots of things are,
> > > like the savings in your bank account.
> > Exponential growth of compound interest is a textbook case
> > where your numerical model of physical layer processes and reality
> > increasingly diverge, requiring periodic, painful readjustments.
> I have never heard of a case where a bank simply refused to pay interest
You have never heard of banks going broke, assets seized, currency
Are you honestly believing that money likes to work, and it
keeps growing in the bank vaults, like early miners thought
metal grew in the mountain, so that they left there some so
that it could breed?
> because there was just too much money in the account. So what are you
> referring to here?
> > > > People
> > > > forget that hard drives stopped doubling, at least for a short while.
> > > >
> > >
> > > Because of a flood in Thailand. Nobody has said there wouldn't be bumps
> > in
> > Thailand was not the reason.
> > We're stuck at 4 TB because they ran into limits of a particular
> > technology.
> I'm baffled by your use of the word "stuck" here. We just got to 4 Tbytes
You're getting far too frequently baffled for my liking. I'm showing
you instances where reality deviates from the nice linear semi-log.
There have been multiple smooth technology handovers in the platter
areal density which however shew a different scaling
If you think you're seeing a linear semilog plot in there, then
throw away your ruler. Or buy new glasses. If you're now agreeing
that the growth is saturating, then why are you wasting my time?
> not that long ago. We always get "stuck" by this definition. I have
> attached my spreadsheet of hard drive prices that I have been maintaining
The metric you're looking for is areal density.
> for a few years, but initially got elsewhere. I welcome comments.
> > In case of platters full of spinning rust the snag is temporary,
> > as there are two successor technologies about to enter the
> > marketplace (HAMR and BPM, not new but close to becoming
> > mature enough for practical applications) so there's probably
> > another order of magnitude still to go before end of the line.
> > That makes it 40 TB.
> That hardly seems like "stuck" to me. Knowing how we're going to get the
> next order of magnitude is good enough for me.
No, deviations from linear semi-log plot are definitely not good enough
for Ray, and I have to agree with him. If you like leaning out of the
windows very far, prepare to deal with gravity.
> Aside from that, there are things out there that promise to give the next
I'm not interested in promises. I'm interested in past data that
show where real world disagreed with prior predictions.
> order of magnitude after that... such as:
> Which is clearly not ready for prime time, but is a good idea of the sorts
> of things that might happen when brilliant people are tasked with an
Brilliant people can't make features smaller than atoms. Brilliant
people have no magic wands to short-cut technology maturation times
so that they spring full-formed from nothing, just as Athena sprang
from Zeus' forehead. Brilliant people can't make a refinery grow
out overnight, for free.
> Coincidentally, NOR flash has recently also entered
> > scaling limits.
> > The time for surface scaling is running out. The only
> > alternative is 3d volume integration. We do not have anything
> > in the pipeline to arrive in time, so there will be a gap.
> > The only technology to interpolate would be Langmuir-Blodgett
> > serial layer deposition, with according 2d liquid mosaic
> > self-assembly/alignment. I'm not aware of this technology
> > to be ready for deployment. Next after that is 3d crystal
> > self-assembly from solution. This is even further away.
> That's ok, we have time to get this stuff right before falling off the
We've already fallen from the semiconductor litho curve.
See the NOR flash scaling at the URL I posted earlier.
We've already fallen of the PV deployment curve (and we
were never on the according infrastructure curve in the
You can't jump from zero TW to 20 TW in 20 years.
Not unless you have MNT, and collectively we made sure we
failed to develop that.
> > > the road, just that there was an overall trend.
> > >
> > >
> > > > People are unaware of finer points like
> > > > http://postbiota.org/pipermail/tt/2013-October/014179.html
> > >
> > >
> > > Ok, I read that, and what it said in a nut shell is "fuck this is hard".
> > Yes, this is the nature of limits. Instead of constant doubling
> > times the last few show longer and longer steps. As I told you,
> > we're no longer at 18 months but at 3 years doubling time this
> > moment. The next doubling times will be longer. This means
> > that linear semilog plot is no longer linear. No more Moore for you.
> And yet, it is still doubling rapidly. The end result is the same, just at
Which part of "no more constant doubling times for you" you don't understand?
> a slightly different time scale. And there is no guarantee that we won't
Which part of "you can't make widgets smaller than single atoms" you don't
> make a hop with a new technology and get back on any given curve. It can
A gold meteorite can fall in my garden. Hey, it could happen.
> > > Not, I expect it to come to a screeching halt.
> > Why do you expect that? Look at the price tag of the
> > zEnterprise 196. Obviously, a somewhat higher margin
> > than on a 50 USD ARM SoC.
> Sorry, you've lost me here. I don't know what these things are.
It's a cheap mainframe, 75 kUSD entry level. Obviously, CPUs
build from such can be made from unobtainium. But flash drives
and mobile CPUs have low margins, so there's diminished incentive
to go to the next node (especially if the next node has lower
performance than current one).
> > > If they read your posts here Eugen, they might decide not to thaw you
> > out.
> > > Who needs a pessimist in a utopia... :-)
> > Utopia? I'm afraid I have another piece of bad news for you.
> > Very bad news, I'm afraid...
> Anyone looking at 2013 from the time frame of 1913 would clearly call this
Anyone looking at an arbitrary time frame knows that darwinian
evolution still applies.
> utopia, at least from the technological standpoint. Also from the number of
> people operating under democracy, decreased violence and a number of other
> points. Not that it is utopia in every way.
My thesis is that a postecosystem has a food web.
> > > > > I know you MUST believe that computers will continue to get faster,
> > even
> > > > if
> > > > > they don't quite keep up the doubling pace. Right?
> > > >
> > > > What does faster mean? The only reliable way is a benchmark.
> > > > The only really relevant benchmark is the one that runs your
> > > > problem. As such faster never kept up with Moore's law.
> > > > Moore's law is about affordable transistors, and that one
> > > > gives you a canvas for your potential. So your gains are less
> > > > than Moore, and sometimes a lot less. For me personally,
> > > > the bottleneck in classical (GPGPU is classical) computers
> > > > is worst-case memory bandwidth. Which is remarkably sucky,
> > > > if you look it up.
> > > >
> > >
> > > The problem I care about the most is computer vision. We are now
> > Computer vision is very easy, actually, and quite well understood.
> You are clearly stark raving MAD. There is no computer on earth that can
> tell a cat from a dog reliably at this point.
There is a significant difference between "I have no idea how to do that"
from "ok, it's mere engineering at this point".
We're understanding processing the retina sufficiently to produce
code that the second processing pipeline can use. We have mapped
features of later processing stages to the point that we know what
you're looking at, or what you're dreaming of. We have off the
shelf machine vision systems for many industrial tasks. We have
autonomous cars that drive better than people.
This is obviusly one of these cases where we've made some slight
progress over last few decades.
> > The low number of layers and connectivity (fanout), all local at
> > that, a retina needs are within the envelope of silicon fabrication.
> The retina is not what I'm talking about. I'm discussing image
But the retina is what I'm talking about, because it's structurally
and functionally simple enough so that Moravec picked it for his analysis.
> understanding. "That is a picture of a dog in front of a house. The house
> has a victorian architecture. The 1957 Cadillac next to the house would
> indicate that the picture was most likely taken between 1956 and 1976."
This is beyond machine vision. This is halfway to human-equivalent AI.
> > > approaching automated vehicles becoming a reality. I thought it would
> > > happen in 2014 since 2004. It may be delayed a year or two by bureaucrats
> > > and lawyers, but the technology should be cheap enough for luxury cars to
> > I'm afraid luxury something is going to be a very, very small market
> > in the coming decades.
> Stop. This is just irritating and unhelpful.
You can bet this is fucking irritating, because we did it to ourselves!
A pathetic failure of planning.
> > I agree that autonomous cars are mostly a very good thing, unless
> > you happen to be a trucker, or a car maker.
> I'm not sure how autonomous cars are bad for car makers. I do get why they
The duty cycle of a personal car is terrible. With autonomous cars you
only need a small fraction of total fleet. You already see the beginning
of this with carsharing smartphone apps. Now you no longer need the hassle
of owning a car just that you can use it. It's just great, unless you're
in the car making business. It's pretty awful if car making is your
country's main moneymaking business, and they're really EV and autonomous
> are bad for truckers.
> > Whatever Germany
> > earns on car making is about enough to pay for the fossil fuel
> > imports.
> You're confusing me again.
Obviously, Germany has to figure out some other way to pay for their
fossil fuel imports in the near future.
> > > have highway cruise control (including steering) by 2014 or 2015. So my
> > > venture into guessing the future was pretty close, using Ray's technique.
> > >
> > >
> > > > Ok. So, now your transistor budget no long doubles in
> > > > constant time, but that time keeps increasing. It's roughly
> > > > three years by end of this year, no longer 18 months.
> > > > Physical feature size limits are close behind, and your
> > > > Si real state is a 400 mm pizza, max. WSI gives you a
> > > > factor of two by making yield quantitative, but it wrecks
> > > > havoc to your computation model, because grain size starts
> > > > being tiny (less than mm^2), and asks for asynchronous
> > > > shared-nothing, and did I mention fine-grained? So no
> > > > TBytes of RAM for your LUT. The next step is FPGA, as in
> > > > runtime reconfigurable. That *might* give you another
> > > > factor of 2, or maybe even 4. Stacking is off-Moore, but
> > > > it will do something, particularly giving cache-like
> > > > access to your RAM, as long as it's few 10 MBytes max.
> > > >
> > >
> > > I've predicted that they will go to 3D. It is the only logical way to go
> > Everybody and his dog predicted that, since early 1970s.
> > The difficult is actually making it happen, just in time
> > when semiconductor photolitho just runs out of steam.
> > Guess what, that time is now. So, where is your 3d integration
> > technology?
> > > from here, other than maybe 2 1/2 D first...
> > You can't have that by semiconductor photolitho. Stacking is
> > off-More. What else have you got?
> > >
> > > > And then you have to go real 3d, or else there's gap.
> > > >
> > >
> > > True, unless something completely different comes along, which may not be
> > > highly likely.
> > New technologies typically take decades of development, until
> > they're sufficiently matured so that they can take on mature
> > technologies that have ran into their scaling limits.
> I totally agree that the 2d processes we are currently using are running
> into limits. But we will keep making the stuff we're making now cheaper. In
Older processes are cheaper than bleeding edge, but that curve saturates
almost immediately. The only way to drop the costs is by using a new technology.
> my mind, that keeps us on Moore until such time as a 3D solution is worked
We are already off-Moore. The question is how long it will take until
a different technology can pick up scaling, at least for a brief while
(if you're at atomic limits in the surface, you're only a few doublings
away from where your only option is to start doubling the volume).
> out that makes things faster.
> The main problem in my mind isn't making stuff smaller, but in dissipating
> heat so you can stack it up close to each other. That's what I mean by 2.5
Stacking is off-Moore. One of the scaling limits is that the power scaling
is no longer with us. As you'll notice, no novolatile memory is ready to
pick up the torch of SRAM/DRAM/NOR flash, despite many decades of development.
> > >
> > > > My guess the gap is somewhere 15-20 years long, but
> > > > we've got maybe 10 years until saturation curve is pretty
> > > > damn flat.
> > > >
> > >
> > > Ok. Then we can start making larger structures. It won't speed up due to
> > > decreasing transistor size, but it will be able to do useful work.
> > Imagine
> > > a 3d CPU 5 inches on a side. That could do some serious work. More than a
> > > human brain.
> > The human brain is a 3d integrated assembly of computational
> > elements which are built from features on nm scale.
> Sure is. And I'm convinced that we'll soon enough have similar
> computational devices. Maybe we'll even figure out how to grow them. Who
> > > > He implicitly implied we'll run on 100% of thin-film PV in 16 years.
> > > > That was 2011, so make that 14 years. This means 4.2 TWp/year just
> > > > for power in a linear model, nevermind matching synfuel capability
> > > > (try doubling that, after all is sung and done -- 8 shiny TWp/year).
> > > > We're not getting the linear model. In fact, we arguably sublinear,
> > > > see
> > > >
> > http://cleantechnica.com/2013/10/14/third-quarter-2013-solar-pv-installations-reach-9-gw/
> > >
> > >
> > > You obviously don't understand the nature of his prediction. If he says
> > Obviously. I expect prediction to be brittle, and that the originator
> > is prepared to eat some crow, in case she is wrong. I'm old-fashioned
> > this way.
> No Eugen. Let's simplify. If I predict that there will be 20 billion cats
> running loose in Germany in 16 years and that the number of cats will
> increase exponentially with a doubling every two years, and you come back
> in two years and say "We don't yet have 2 billion cats, so the prediction
> must be wrong." Then you clearly don't have an understanding of what
> exponential means.
I'm sorry, this is too stupid, I'm done with this mail.
This is the kind of argumentation that made me stop talking to John.
> > > that the doubling in solar efficiency is 3.5 years (going from memory)
> > then
> > > half of the solar he envisions will be installed between July 2023 and
> > Thank you for explaining exponential growth to me. I think I've first
> > understood it before I was 10.
> Clearly not so much.
> > The nature of solar cells is the only
> > way to double the output is to double the surface. And the according
> > infrastructure in the background, simple things like 10 GUSD plants,
> > electric grid upgrades, storage systems, and the like.
> > > 2027. What's being installed now probably is sublinear, that's what an
> > > exponential would predict. He didn't predict a linear model. We'll
> > revisit
> > You're not understanding me. It used to be exponential. Because it's
> > very easy to double very little. Until suddenly you have to double
> > quite a lot. This isn't a lily pond or a bacterial culture, this is
> > infrastructure work.
> > So Ray is already wrong, once again. The trend is no longer exponential.
> The only thing Ray said about solar is that the cost of the panels
> themselves was on an exponential curve. There is a lot more to solar costs
> than that. So if you hold him to something he didn't say, then yes, we're
> off the curve. If you have data showing that the panels themselves have
> fallen off the curve, then I'll give you credit for being correct, and will
> admit that Kurzweil is off, at least for the moment. (Unfortunately for
> you, you can sometimes hop back on the curve later with a new technology.)
> > > his prediction in 2027 if we're both still communicating by then.
> > The prediction is 100% of electricity in 16 years. He then scaled
> > that back by saying 20 years. That's 2021.
> > Given that we're already off-exponential, I expect that you keep
> > posting "I'm wrong" every year.
> I will give you this. Ray made a prediction about the price of solar panels
> and then extrapolated that to the price of solar generated electricity
> (which includes infrastructure, inverters, batteries and the like) which is
> not the same thing. I don't think Ray is right when he says things that
> imply that lots of infrastructure will just pop into existence. But, how
> much more infrastructure would pop into existence if solar panels were
> nearly free? A little more, but the other parts are still damn pricey,
> better than half of a current home installation.
> > > > apply. However, in a sense it does apply. We do get some percent
> > better at
> > > > > extracting what's left each year. That doesn't mean we get an
> > exponential
> > > >
> > > > No, in terms of net energy we're not getting better. We're actually
> > > > getting worse.
> > > >
> > > > > amount of oil, since there's a limited amount of the stuff. But it
> > does
> > > > > mean that we get exponentially better at finding what's left (note
> > that
> > > >
> > > > We're not getting better. We've mapped all the stuff, there are almost
> > > > no unknown unknowns. And dropping EROEI and even dropping volume
> > > > (not net energy, volume!) per unit of effort is pretty much the
> > > > opposite of exponential. Do 40% of decay rate/well/year mean a
> > > > thing to you?
> > > >
> > >
> > > You misunderstand my point again. I know it's harder to get oil. But we
> > > develop new technologies for getting at what's left.
> > Fracking is 40 years old. Fracking is running into diminishing returns.
> > So where are your new technologies, which need to be already in wide
> > deployment, now?
> Fracking is currently producing a fair amount of oil. It's produced a lot
> of buzz. I am not an expert on oil production techniques, so I'm not going
> to argue from ignorance.
> > > > > this curve is likely much more gentle than computing, with a
> > doubling of
> > > > > reserves we can get at maybe every 20 or 50 years. I don't know.)
> > > >
> > > > There are no exponentials in infrastructure. There is an early
> > > > sigmoidal that looks that way, but we've left that already.
> > > >
> > >
> > > Infrastructure can change rapidly. How long did it take for everyone to
> > get
> > No. Infrastructure takes 30 years, frequently longer. That's a constant.
> > > a cell phone? Smart Phones? When electric cars make financial sense (if
> > How long did take for everybody to get their own synfuel plant?
> I agree larger infrastructure is harder. But things can change rapidly when
> they have to.
> > > they ever do) then people will switch to them quickly. Large
> > infrastructure
> > Do you understand the logictics of car production? Battery manufacturing?
> > Dynamics of fleet exchange? Recharging infrastructure? Including the money
> > to fund it all? Do the math, it is really quite illuminating.
> I'm sure it is a hard problem. I do know the US car fleet turns over every
> 16 years or so. That would indicate that if it became a real issue and we
> had a real solution that it could be done in 16 years.
> > > like roads and so forth will remain problematic until robotics is good
> > > enough to do much more of the job.
> > > > >
> > > > > > This is the opposite of science.
> > > > > >
> > > > >
> > > > > It is a part of science, the hypothesis part. LAR applied to
> > computing
> > > > > available per dollar in particular is a hypothesis formed in the mid
> > > > 1960s.
> > > > > As far as I know, we are still more or less on that track, though
> > they
> > > > have
> > > >
> > > > No, we're not. See benchmarks.
> > > >
> > >
> > > Data please. I can't find any. I have looked.
> > Try Stream, though it's a synthetic benchmark
> > http://www.admin-magazine.com/HPC/Articles/Finding-Memory-Bottlenecks-with-Stream
> > It would be a reasonable assumption for retina-like processing
> > scaling. Deeper visual pipelines are different. Here, you need
> > to access something like fetching from a large (>>10 GByte) of
> > random pointers.
> A coprocessor with direct access to the "memory" of a retina would be very
> good for problems like this.
> > > > People who hear about Amdahl's Law the first time have to stop
> > worrying,
> > > > and embrace nondeterminism. People who expect reliable systems at
> > hardware
> > > > level are gonna have a bad time.
> > > >
> > >
> > > I disagree with that. There will be reliable hardware, or they won't be
> > If you want to not run into Amdahl you need to embrace nondeterminism.
> > Building test harnesses just got a bit harder.
> You would need a new generation of programmers to go there, I think.
> > > able to sell it. No matter how slow the previous generation was. It is
> > hard
> > Yes, there will be unreliable hardware. This is one of the problems in
> > exascale: unreliable transport and unreliable components (as in: parts
> > of your system keep failing at runtime, and you diagnose and remap to
> > hot spares, all without breaking a stride). Beyond that, you've got
> > stochastical computing elements. That's one of the joys of living at
> > nanoscale.
> But it has to be dealt with at a low level, or programmers won't stand for
> > > enough to get programmers to do multi-threading. It would be damn close
> > to
> > > impossible to get them to switch to a model where the answer might not be
> > > right.
> > There is no longer "exactly right" there is only "good enough".
> Well, for now we have exactly right for computation, in the future, and for
> certain algorithms that are NP complete, for example, good enough is fine.
> But for credit card processing, good enough isn't good enough.
> > > A pessimist will just hole up in his cave. I refer you to "The Croods" to
> > I don't know what a pessimist would do. I do know that the only guy who'd
> > still have water when his cars break down in the desert is a realist.
> > The optimists always end up as bleached bones. Your call.
> I prefer the extra years of life. Perhaps I'll spend them in a cave in the
> extropy-chat mailing list
> extropy-chat at lists.extropy.org
More information about the extropy-chat