obviating the need for physical experimentation? bah, we'll obviate the need for physical reality. think big!

The other day my friend Gary emailed a link to me for an article on this computer. I am normally sceptical of such talk, too, Barron but the asymptotic curve is phenomenal. Computing technology has experienced a logarthmic leap in the last generation and it shows no sign of slowing. Moore's Law will be tapped out in a few decades, though, once the atomic level is reached. There already is research underway on computing with atoms, which is quantum computing - instead of 1 or 0, you have both 1 and 0, simultaneously and the possibilities are mind-boggling, if the technology can be made practical.

Super cool post Barron. Amazing news, even more amazing graph.

I think people are in a state of total confusion about the future. On the one hand Brad likes to depress us with his destruction of all possible energy alternative arguments and relishes every opportunity to point out that the political will isn't there, even if there was a solution. We then get hit with dire predictions of the burning earth hell we'll create if the planet heats up just one degree.

On the other hand the news in advances in biotech, nano and this stuff isn't impressive, it's almost impossible to comprehend.

At times it seems as if it really is a race between our greed and stupidity and our ingenuity and intelligence. Or Conservative v Liberal as we know it today.

I just hope I live long enough to see Call of Duty 20. Wow, COD 20 on a petaflop Xbox!

I hate to rain on anyone's parade, but this accomplishment is not quite as amazing as it might seem. The petaflop of computation offered by this machine can only be exploited by *very* specialized programs, which take huge engineering effort to create. Furthermore, many programs can't (currently) be written to effectively use that processing power.

Why? The key lies in parallelism. RoadRunner does not achieve a petaflop with one, or even just a handful, of chips. It uses about 20,000 of them. Each of these chips (or cores, as they are now called), run at modest speeds, not much different that what's in your laptop or on your desktop. When you put them together, though, they add up quite nicely.  However, that requires that your program can be broken up into 20,000 little pieces that can all be run in parallel and then have their results efficiently and effectively sewn back together. Unfortunately, that's really hard to do for most tasks. For any task in which step 2 (or 3 or four ...) depends on step 1, you have to wait until step 1 is done to run step 2.  So, you can't run such steps in parallel. Furthermore, even if there are many steps that don't depend on each other, and so can be run in parallel on different chips, they have to be big enough to be worth the cost of starting them on different chips and then collecting their results.

The good news is that many scientific applications fit this descriptions, particularly simulation of physical reality, which often involves a great deal of truly (or mostly) independent activity. The bad news is that writing scientific applications this way is *really* hard (mere mortals need not apply) and plenty of other applications don't fit this description.

What about Moore's Law (mentioned by one poster)? Unfortunately, Moore's Law hit a brick wall a few years back.  The power needed by a chip is proportional to something like the square of the clock speed. So, today's chips are using a lot of power to run. That power results in a  *lot* of heat, which needs to be taken away from the chip to prevent it from melting. However, within the past few years, chip designers realized that the clock speeds are pushing the limit of their ability to effectively cool the chip. Hence the lack of significant speed increases over the past 5 years.  (A New York Times article from a few days ago mentions this issue, you can read some more details there and in articles it links to:

bits.blogs.nytimes.com/.../apple-in-parallel-turning-the-pc-world-upside-down)

What's a chip designer to do? Well, if you can't get the chip to go faster, why not just sell the customers more chips? Hence, dual-core, quad-core, ... chips. Simultaneously, you spend lots of money supporting research into how to actually *use* this chips effectively. So, Intel for instance, refocused its Labs on this problem a number of years ago already, and are actively supporting academic research in this area. Still, there have yet to be any breakthroughs in this area.

Can it be done? That is, can we at some point figure out how to use these chips effectively?  Perhaps. The previous surge of research into parallel programming pretty much flopped. But, back then, people still believed in Moore's Law. Now that industry and research have realized the reality, serious efforts are underway to find innovative ways to effectively use massively parallel machines (think Intel Core-128 as a standard desktop chip within a few years).

To sum up -- yes, there's reason to be excited. However, the reality is that the computing world is actually facing a crisis the likes of which it hasn't seen before, namely, the end of Moore's Law. Simultaneously, undergrad enrollment in Computer Science in college has dropped precipitously since ~1999, government funding of all areas of Computer Science, including those most closely addressing this issue, is drying up, and there's no end in sight. Industry is for the large part *not* stepping in to make up the difference. So, in response to the Ignorant Populist -- even technology can't beat politics. And for Barron, if the US goverment doesn't do something about the current funding situation, then pretty soon such posts will end in "And it was all invented there, in good old <fill in foreign country here>".

"Total confusion about the future" is right, Iggy Pop.   It is a race between our better and our worse selves.  And this relates back to the 'Are Environmentalists Getting in the Way' post.  It's clear--to me anyway--that success will be the result of bigger and better tech like these ultra-fast computers and the oilgae mentioned here last week, not through reversion to some simpler subsistence farming past.  Too many environmentalists waste all their thought and energy romanticizing the wilderness and our hunter-gatherer past.  They fail to recognize that the only way we'd get back to such a "balanced" form of existence is through a massive die-off,  95% of the current world population, say.

We are approaching a singularity, several of them, in fact.  We're farther back on the shoulder of the graph when it comes to fossil fuel use and CO2 emissions, but those too are increasing asymptotically.  And this goes to the reason as to why we're confused about the future.  Mathematically speaking, a singularity is undefined.  When you have several relevant curves rearing up towards infinity--this in an objectively finite system, i.e. the world--it becomes difficult (impossible?) to predict what's going to break first.

That's not an asymptotic curve, libref.  An asymptotic curve would never reach one petaflop ... or never reach 2008.  That's just a curve with a high rate of change recently.

Also, excellent post yrmandel.

aeromonas...you're absolutely right that we're reaching the point of technological singularity in many areas. I'd say that we're reaching a point of criticality in human capacity. I am sure that with the increase breakthroughs in material technologies, applied nanotechnology, applied material efficiency (were a material does more than one thing, think gore-tex here) and the computer breakthroughs as far as computing power and material miniaturization.

You're point about " Too many environmentalists waste all their thought and energy romanticizing the wilderness and our hunter-gatherer past" is partly true but I'd say that many contemporary environmentalists and by 'contemporary" I mean those that are actively looking for applicable solutions using current technologies to make the human impact less negative while still maintaining or exceeding quality of life for humans and the natural environment. There is always going to be a sliver of any population that waxes poetic about a pastoral past they've never lived nor experienced whether it's living in the wilds of Alaskan frontier or reverting to cultural revisionism via social reductivism.

As much as I might come off as pessimist on some of these environmental issues, I'm really a pessimistic optimist. I think that we can figure a way out of all the issues and problems caused by human activity if we start forward thinking.

As I've posted before on other areas is that for civilization to succeed in the near and long term future we're going to have to think in 21st century terms and not 20th century terms when we rethink and develop solutions to pollution, energy, waste, smart materials, etc and even revisiting the idea of taking cues from nature itself. We also need to share these technologies worldwide because it allows developing 3rd and 2nd world nations to leapfrog the slow-climb out of poverty by taking advantage of 21st century thinking instead of insisting they reinvent the wheel all over again.

To play off of singlespeed's post, I would generally agree that there is a split in the green community.  There are the people who are environmentalists for nature's sake, and those who are environmentalists for humanity's sake.  The first group might say "We need to stop global warming because it threatens the habitat of polar bears!"  The second might say "We need to stop global warming because warmer air leads to less snow on mountaintops, which means there will be less water in the rivers downstream to grow crops for us to eat!"  Unfortunately, many of the pleas that get into the mainstream are more with the first kind, to which the response is, not surprisingly, something akin to "Why should I care about polar bears?  I've gone long enough without seeing them, and I don't think I'm a worse person because of it.  I don't have a hole in my soul because I haven't seen any dodos or giant ground sloths.  I won't change my life to save something that, however cute, has no bearing on my daily life."  To have any relevance, green arguments must be made couched in language that describes environmentalism as sustaining humanity, not sustaining nature; sustaining US, not sustaining IT/THEM.

...and to make this post relevant to the thread, yay for microchips!

Iggy says, "At times it seems as if it really is a race between our greed and stupidity and our ingenuity and intelligence," and describes human history.

yrmandel:  So, someone here has relevant knowledge regarding the topic and wants to share it with others in a clear, coherent fashion.  There's only one word for that: elitist.

yr:  A question:  Can you advise as to when we will have computers implanted in our heads controlled with brain signals which will also include communications components enabling us to access the world's knowledge and talk to others without external devices and in a totally seamless way.  My basic problem is that I'm running out of places to put all my junk.  I could quit smoking, but screw that.  I also want to be able to take pictures with my eyes.  I'll take my answer off the air...

To give another perspective.

30 years ago the fastest computer in the world was a Cray 1-S which ran at 85MHz and had 8MB of RAM. My cellphone has a 400 MHz chip and 2GB of flash memory.

yrmandel writes:

-- The good news is that many scientific applications fit this descriptions, particularly simulation of physical reality, which often involves a great deal of truly (or mostly) independent activity. The bad news is that writing scientific applications this way is *really* hard (mere mortals need not apply) and plenty of other applications don't fit this description.

The first sentence is true, the second sentence is false. Some time ago I used a supercomputer with 4096 processors and didn't notice any increase in difficulty over using a supercomputer with a single processor. It required different algorithms but learning them was certainly not *really* hard.

yrmandel continues:

-- Well, if you can't get the chip to go faster, why not just sell the customers more chips? Hence, dual-core, quad-core, ... chips. Simultaneously, you spend lots of money supporting research into how to actually *use* this chips effectively. So, Intel for instance, refocused its Labs on this problem a number of years ago already, and are actively supporting academic research in this area. Still, there have yet to be any breakthroughs in this area.

No breakthroughs!? whiskey tango foxtrot. I would not be surprised if most people reading this blog are using two-core computers. Intel spends a lot of effort making that happen as do Microsoft and Apple - but the further up the technology chain you go the less you need to care about the number of cores.

In 2007, John Bolton wrote that Republicans had achieved &quot;the end of arms control.&quot; He was