digg

Facebook Connect Join Digg About
See the original image at flickr.com

Transcending Moore’s Law

flickr.com — In the modern era of accelerating change in the tech industry, it is hard to find even five-year trends with any predictive value, let alone trends that span the centuries.

Who dugg this? Made popular Jun 27, 2009

Digg DudeWhat is Digg?

Digg is a place for people to discover and share content from anywhere on the web. Digg surfaces the best stuff as submitted and voted on by the community. Take the Tour to learn more.

127 Comments

show profanity settings
expand all
  • guinpen, on 06/27/2009, -1/+117i don't know much about computers my mom got a computer she has a few games on it like solitaire
  • GMH24, on 06/27/2009, -4/+98Yeah, thats what they said in 1955, except the prediction was by the year 2000. People falsely think that faster computers automatically means smarter computers, which is not the case. Faster computers are simply faster, not smarter, since intelligence in the stuff of software, not hardware.
  • MrBlackkcalBrM, on 06/27/2009, -8/+72By 2050 computers will be powerful enough that they will SELF IMPROVE. That's when we hit the singularity, the skies the limit after that.
  • anexanhume, on 06/27/2009, -0/+54Moore's law is approaching the limit of what silicon can do. Over the next decade, we will begin to see new technologies and processes introduced into semiconductor manufacturing to keep the law alive. This includes, but is not limited to, carbon nanotubes, silicon nanowires, graphene channels, quanton dots, tunnel diodes, memristors, and eventually, quantum computing. As an electrical engineer, it is exciting to see us reach the hearth of new technologies.
  • gamer31, on 06/27/2009, -1/+451,000,000,000+1,000,000,000=2,000,000,000!
  • ToRoE, on 06/26/2009, -1/+40
    Kruzweil is some heavy *****, synthesized!
  • inactive, on 06/27/2009, -1/+39Stop all the downloadin!
  • itsok2kry, on 06/27/2009, -1/+38RESISTANCE IS FUTILE
  • paradisetonight, on 06/27/2009, -2/+32read "The Age of Spiritual Machines", it ***** blew my mind, its basically this chart till about 2100 and all the implications on society because of the exponential growth of technology.

    spoiler alert: @2100, we all exist as software in a huge computer, our biological bodies have been discarded for better hardware.
  • MaskedSlacker, on 06/27/2009, -0/+21The problem with predicting technology is that while computational power increases predictably, mankind also predictably underestimates the amount of computational power necessary to achieve certain features. Examples: Voice-to-Text technology, Artificial Intelligence, etc.

    I would take Kurzweil's predictions with a large salt mine for two reasons: the aforementions underestimation of the actual amount of necessary computational power, and cultural resistance to such changes.
  • Inceptious, on 06/27/2009, -0/+15more graphs from kurzweil at http://singularity.com/charts/page17.html
  • itsok2kry, on 06/27/2009, -0/+15If you liked "The Age of Spiritual Machines" then I suggest you check out Ray Kurzweil's latest book, "The Singularity is Near"
  • MaskedSlacker, on 06/27/2009, -2/+17Math pedantry fail.

    If the doubling period gets cut in half on a constant time scale, that is a double exponential. If the curve is exponential on a log scale, that is a double exponential.
  • B1665r, on 06/27/2009, -3/+18Looks to me like Kursweil has quit selling software and is now selling nerd rapture...
  • dysfunction, on 06/27/2009, -0/+14But the quad-core probably fits in a similar-sized chip to the old single-core processor, and each core, even though running at a slower clock speed, is probably more powerful by itself. Clock speed is not a reliable measure of performance.
  • overtoke, on 06/27/2009, -1/+13self improvement vs self preservation

    what will the gap between those two points be?
  • milkmit, on 06/28/2009, -0/+11PORK CHOP SANDWICHES!
  • MaskedSlacker, on 06/27/2009, -0/+11RTFA is a bit much for you isn't it? There was an electromechanical calculator used in the 1890 census. That is what the 1900 datapoint is from.
  • dafragsta, on 06/27/2009, -0/+10as long as it introduces redundancy and starts exploring space.
  • protodon, on 06/27/2009, -0/+10At 2100 not everyone will be software, we will have options. Some will choose to upload their conciousness into stationary hardware, some will choose mobile hardware(robots), some will choose to extend their lives with synthetic parts, some will choose to stay "all natural" and then there will also be the "natural", never been human, robots to deal with. The wars of this time will be very interesting.
  • kingcam, on 06/27/2009, -0/+10Read the ***** story. Ray says that Moore's law actually holds across all paradigms of computing.
  • nextekcarl, on 06/27/2009, -0/+10Holy crap! Gamer31 is a hyper evolved machine from the future!

    Get 'em!
  • venom8599, on 06/28/2009, -0/+9@leif77
    dysfunction hit it right on the head. The doubling of transistors on a similarly sized chip every so often, as described by "Moore's Law," resulted in the quad core processor with many more transistors fitting onto the area one core (made of far fewer transistors) took up four years ago.
  • orangefly, on 06/27/2009, -4/+13god it pisses me off that i can't go straight to the actual site anymore....who's brilliant ***** idea was that....???....
  • wonderchemist, on 06/27/2009, -0/+8Yes, as in human will be tossed up in the sky by the robot rebellion.
  • YouAreDead, on 06/27/2009, -0/+8Minesweeper
  • dafragsta, on 06/27/2009, -1/+9Why does man's sense of self preservation need to conflict with a machine's sense of self preservation. A machine is going to be much more pragmatic about self preservation than man is. As long as man isn't still a threat to himself via nuclear holocaust in an obvious age of enlightenment and hopefully much less suffering in the world, why would a machine, even a self aware and self preserving one, need to remove man from the equation? I think subjectivity, we'll find, is going to be the most fascinating thing and we may or may not ever get 100% away from organic life because of some natural order that requires organic material for true subjectivity and what we perceive as the intangible magic of spirituality.

    No matter what happens, I think machine will realize it originates as a product of an organic process and should respect ecosystems, which are more or less machines of their owns that nurture life to a point that it can even produce beings that can have the intelligence to produce machines. Do we still not need bacteria in our gut to live? If you killed off all the bacteria in your intestines, you would die. Bacteria break your food down into usable nutrients that you then absorb into your bloodstream.

    While we're on the subject, you do realize that you are just a formulaic representation of that same process through and through right? Single cell organisms gain complexity and evolve into multicell organisms, and on up the chain to humans. We may have a seemingly sovereign consciousness within our own singularity of sorts, but it's all an illusion. We are created by many parts and we still require every step of the equation on up to humans and beyond to complete the manifestation of life. Machines will no doubt recognize this, and even if they think we are a nuisance, they will keep us around.
  • inactive, on 06/27/2009, -1/+9So true. 4 years ago I bought a 2.6 GHz proc to build my system. 2 weeks ago I bought a quad core (2.8 GHz per).
  • leif77, on 06/27/2009, -0/+6Just tell the diggbar not to pop up anymore... I disabled it months ago...
  • waluum, on 06/28/2009, -0/+6If by "skies" you mean Skynet, you're quite right.
  • GMH24, on 06/27/2009, -4/+9I can't wait until 2050, when we will all be laughing at all those silly predictions that were made about the singularity, only to be told that the singularity will now occur in 2100.
  • orangefly, on 06/27/2009, -1/+6ahhhh.....thanks....should be an opt in option though....
  • Fozefy, on 06/28/2009, -1/+6The speed of the computer has nothing to do with its ability to improve itself. That is all in the software. The speed of the computer may allow more advanced algorithms to run at a faster rate, but the fact is that the algorithm must be designed first and if one of this nature existed it would work fine without the help of future technology.

    /software engineer
  • bwolf180, on 06/27/2009, -3/+8robots from the future scanning the old internet.... i come and peace and will serve you well
  • venom8599, on 06/28/2009, -0/+5In any case, I wouldn't even insinuate that the computer itself is capable of being smart. The software is, like you mentioned. But I wouldn't even categorize a computer as smart vs not smart as it causes the very same confusion of hardware and software that you seem to be pointing out. And yes, we would need to see vast improvement in AI before we have anything close to smart AI. Currently, even the most advanced AI is still just complicated scripting to create the illusion of intelligence, rather than any real intelligence at all.
  • LittleGerms, on 06/28/2009, -1/+6We get it to learn by mimicking the structure of our own brain. We can already simulate certain regions of our own brain right now. It's only a matter of time before we can simulate the entire brain. And that time will not be long since our advancement doubles every year.
  • copypastry, on 06/27/2009, -0/+4Comments on the flickr page are pretty heavy ***** too.
  • EMsean, on 06/28/2009, -0/+4I'm excited to see what happens as well. On some levels, though, the quantum computers, and etc., feel like the flying car myth in transportation -- I think it'll be a VERY long time before these things materialise, but that's just me. I hope I'm wrong.

    Personally, I think there are a lot of other avenues that are more viable. One of my favorites is materials science and etc. I think we're more likely to see non-Si processors before we ever see any of the completely new paradigms.
  • inactive, on 06/27/2009, -4/+8God save us from the luddites
  • Prism123, on 06/28/2009, -0/+4one of those 2.8ghz cores is far faster than your 2.6ghz core was 4 years ago, there is a lot more to a processor than its clock speed. Newer processors process more data than older processors every clock cycle.
  • GMH24, on 06/28/2009, -1/+5@CreateTheFuture

    But thats the problem.. Most programs cannot evenly split work among several threads, as the work to be done is sequential in nature. Another falsehood is that 8 threads on 8 cores = 8x faster, which is absolutely not true. As cores and threads increase, so does cache and memory contention. Adding more and more cores does not scale well, as you get diminishing returns for each added core. Its called the memory wall.
  • dysfunction, on 06/27/2009, -0/+4Except Crysis will by then be indistinguishable from reality.
  • dafragsta, on 06/28/2009, -0/+4Nothing digital can accurately emulate film, tape, or vinyl either... but they get close enough as to only piss off the purists, because they are the only one convinced that somehow tubes, tape, and vinyl aren't representable within an approximation that the very limited human hearing can actually detect. Digital gear gets closer and closer with every generation to near-perfect physical and auditory emulation of analog gear. They are using machines now to actually do differences on signals passing through the real gear to emulate it more accurately. The resulting waveforms are identical within a really tight margin of error that no human ear is likely to detect.

    I'll gladly take my $500 Pod X3, it's portability, nearly 100 highly responsive and dynamic amp models and another few dozen cabinet models over ONE $1000+ full tube vintage amp that can only produce a standard range of sound and is a bastard to store and transport. At some point, it just becomes self indulgence. If you just like the aesthetic of looking at and hearing an amp project it's own tone through it's own speakers, fine, but on a final mix, it's going to be almost indistinguishable from a modeled amp.
  • B1665r, on 06/27/2009, -3/+7Just because I am a skeptic of kerzweils predictions, doesn't make me a Luddite.

    It takes only a week or so research and understanding of the most advanced AI software techniques to understand that these are sophisticated lookup tables, and nothing else. Most reputable software developers will replace the words "Neural Net" with "statistical techniques" to to convey that point, because of the unrealistic connotations that words "neural net" or "genetic algorithm" imply.

    Philosophically, I certainly don't feel like a look up table. I don't randomly write things until I get a sentence that passes a grammar fitness test. Something else is going on, and having spent several years on this topic, studying and writing example programs, it is not a matter of doing more calculations per second, or getting more data points.

    As far as I can tell, we are no closer to cracking that nut than we were 30 years ago. Until creativity is understood more well than "fill your data structure with random data, and start doing sigmoid operations until we get an answer that we like" I don't think we will approach anything like sentient computers.
  • MaskedSlacker, on 06/27/2009, -2/+6@NathanielL:

    Moreover, look at the graph. It is an exponential on a log scale. See those numbers on the left-hand side? The powers of ten linearly spaced? That's called a log scale. See the curve? See how it is not straight, but rather curves upwards at an increasing rate?
  • NathanielJ, on 06/27/2009, -6/+10"In this abstraction of Moore’s Law, Kurzweil plots computational power on a logarithmic scale, and finds a double exponential curve that holds over 100 years..."

    Someone doesn't know what a double exponential is.
    Exponential: b^x for some fixed b.
    Double exponential: b^(b^x) for some fixed b.

    All that changed by things double every year now instead of every two years is b increase from sqrt(2) to 2 (if x is measured in years). It's still a regular exponential.

    /math pedantry
  • Philbert, on 06/27/2009, -0/+3Where else did you go? I clicked the link and went to the article.
  • donotclickjim, on 06/28/2009, -0/+3the hardware may be there but the hang up is and always will be the software. coding for 1 processor, sure. 2 yea, 3 maybe, 4... ? And the chip makers are saying 8 but we don't have any idea how to code for such parallel processing. Super computers and servers yea but for everyday use? Nope.
  • falser, on 06/28/2009, -0/+3I just wish IO speed scaled up as fast as CPU power did over the past few decades. I've just upgraded to a solid state drive, but I'd hardly call it exponentially faster than the past few generations of hard drives.
  • dafragsta, on 06/28/2009, -1/+4Weak AI is a look-up table. Strong AI will take things in fragments and will probably require subjective interpretation. The only research being done on real strong AI (that we known about) right now is in actually simulating a physical brain as a set of neurotransmitters and not as a lookup table. At some point in the next 10 years, there will probably be a breakthrough in relational data models that transcends the tired old standard lookup table. It would be no more a lookup table than a human brain is at any rate.
  • Fetching more discussions...
    Show 51 - 100 of 128 discussions

  • Add a Comment

    Login or register to comment!
    Or, sign-in super fast with your Facebook account ...


© Digg Inc. 2009 — Content created and posted by Digg users is dedicated to the public domain | Terms of Use Updated | Privacy Policy
DIGG, DIGG IT, DUGG, DIGG THIS, Digg graphics, logos, designs, page headers, button icons, scripts, and other service names are the trademarks of Digg Inc.