Now That The Exaflood's Debunked, Fear The Exacloud!

from the looming-capacitastrophe dept

Cable and phone company lobbyists (and their army of PR, consultant and think tank friends) have long pushed the bogus concept of an "exaflood," or the idea that explosive Internet growth will result in the Internet collapsing any day now. The argument is generally used by telecom lobbyists to scare politicians and the public into supporting something (deregulation, subsidies, higher prices, fewer consumer protections) lest the Internet explode. The problem is that the argument has been debunked countless times by real network researchers like Dr. Andrew Odlyzko of MINTS -- who highlight that traffic growth is actually quite reasonable, and what growth there is can be easily dealt with by intelligent network engineers and modest network investment. If carriers aren't investing money back into the network, it has nothing to do with bandwidth bogeymen -- it's usually because they face limited competition.

The exaflood term itself was actually coined by Bret Swanson, formerly of the Discovery Institute -- the think tank hired by evangelicals to help push creationism into the classroom via "intelligent design." Under the employ of major carriers, Swanson first used the term in a 2007 Wall Street Journal editorial, and despite it being largely nonsense -- it quickly became a common phrase in modern telecom lexicon. Of course the exaflood never arrived because it doesn't actually exist, but that's not slowing Swanson down. With the FCC considering network neutrality rules, Swanson (now under his his own brand: Entropy Economics) has given the ungracefully-aging exaflood myth a botox injection, based on filings this week with the FCC (via Ars Technica):

"We are intrigued by one particular innovation just around the corner. Call it online gaming. Call it cloud streaming. We call it the "exacloud." It is cloud computing but of a scope and scale never seen before. . . This exacloud will transform video games, movies, virtual worlds, business software, and most other media. Piracy goes away. So do DVDs, game boxes, and maybe even expensive personal computers. New content and software subscription models open up. Based in the cloud instead of on your device, interactivity thrives."

This miracle, piracy-curing super computing evolution Swanson references? It's just ordinary people using clients to access servers using networks. While Swanson throws out a lot of data points in his filing, none of them dispute the reality that Internet traffic growth remains reasonable and manageable. Amusingly, he even goes so far as to use the MINTs data that debunked his original claims -- as evidence supporting his "new" argument. It appears that all he's done is rename his imaginary bandwidth apocalypse for a more modern audience -- and hoped nobody would notice. He at least could have been a little more entertaining. How about the Tubeogeddon? BitTorrentialCollapse? The Tubeacalype? Capacitastrophe? The looming colocaust? Help us out...

Filed Under: bandwidth, exacloud, exaflood, lobbyists, telco

Have We Reached The Limits Of Silicon?

from the are-we-there-yet? dept

The growing demand for greater mobile computing power is rapidly transforming how the semiconductor industry needs to think about designing chips to meet the performance requirements for mobile devices. With GHz-level processor speeds and several GB of data storage, current smartphones have the computational power of high-end desktop computers from just ten years ago. As we move towards smaller and faster computing devices, the use of silicon in chips is making it increasingly difficult to uphold Moore's Law, not only in terms of the laws of physics, but also in terms of economics. However, it's worth mentioning that claims of the end of Moore's Law have been going on for years and are greatly exaggerated when you consider that Moore's Law keeps getting redefined as time passes. In the end, it really isn't as big a deal as the press makes it out to be.

Recently, TSMC's vice president of research and development, Chiang Shang-yi, said that current silicon chip technology should be able to sustain Moore's Law for another decade, but he also pointed out that Moore's Law could come to an end sooner due to economic rather than technological reasons, as it becomes increasingly expensive to develop and manufacture next-generation chips. Even so, he's not really that worried, as his take-home message is that after the end of Moore's Law (whenever that may happen), there are still many more years of what he calls "More Than Moore" technologies, and then even more years of system integration improvements.

Kevin Kelly has presented an interesting and compelling take on Moore's Law -- basically saying that the exponential progress is inevitable in technologies that scale down to microscopic or even nanoscopic sizes. He also points out that once the exponential growth of a technology starts to plateau, we will naturally shift our focus to other alternative technologies, which may subsequently experience their own exponential growth and establish new "laws." He even suggests that Moore's Law could be redefined as a larger trend that can continue indefinitely, encompassing several smaller, overlapping technology trends.

As "the end of Moore's Law" looms ahead, it would appear that people are beginning to shift some of their focus to new alternative technologies that could potentially replace silicon in chips. Here are just a few of the latest technologies that are being investigated for potential use in microelectronics:

In the past couple of years, graphene has been touted as a potential replacement for silicon. Graphene, a single-atom-thick layer of carbon atoms bonded together in a graphitic structure, has been widely studied for its interesting mechanical, chemical, and electronic properties. Graphene sheets have carrier mobilities that are hundreds of times greater than that of silicon, making them ideal for faster chips. However, a major problem with graphene is that it tends to get very hot when devices are operated at the saturation current limits.

Recently, it was demonstrated that memristors -- resistors with memory -- could perform logic operations, and it has even been predicted that memristor-based processors could one day replace the silicon in e-reader displays, as well as in computers.

Diamond, an excellent thermal conductor, can be turned into a semiconductor with the right impurities, and it could be used to make high-performance chips that won't need power-draining cooling systems. However, it is difficult to make diamond wafers large enough for mass production.

Of course, none of these technologies are close to being ready for commercialization yet, but they're a glimpse of what could be the continuation of Moore's Law in its broader sense. In the meantime, companies will try their best to stretch out current technology for as long as it makes sense to.

Filed Under: moore's law, silicon

Innovation By Imitation: Study Shows That Success Comes From Imitation

from the go-with-what-works dept

We've discussed in the past the differences between invention and innovation -- where invention is the creation of something new, and innovation is the actual process of putting it into practice. We've pointed out that the patent system is supposed to encourage the latter (innovation -- as seen in the command that the system "promote the progress") but in practice tends to promote the former at the expense of the latter. The problem is that people who aren't that familiar with the process of innovation think that the two things are the same. But, in reality, innovation is an ongoing process, whereby people have to keep trying out different ideas to make something useful. Anyone who's built a business learns this quickly: the original idea is usually meaningless by the time anything successful comes around. Real innovation is a process of continually trying out new ideas and tweaking them slightly until you figure out what really attracts the market's attention. Studies have shown that real innovation is this kind of ongoing process, rather than the "flash of insight" concept pushed by patent system supporters.

Of course, when innovation is an ongoing process, patents tend to hold back that process. That's because they make it so that only one player in the field -- who perhaps is not the one best equipped to innovate -- gets to run that process. Everyone else is held back. It also slows down the pace of innovation, since without competition, the patent holder has less incentive to keep trying out those new ideas to find what works best. As we've learned for years and years, competition breeds innovation -- but the patent system is designed to cut out competition for no particular reason.

Defenders of the patent system will often claim that the more socially beneficial result is for competitors to come up with something completely new, rather than building off the work of others. However, there is little evidence to support this particular interpretation. In fact, most research into true innovation shows that it is much more efficient for all parties to have access to as many possible solutions as possible -- and holding back those options results in sub-optimal social results.

Yet another new study has shown this to be the case. Some researchers ran a contest of sorts, where they asked various people to submit "programs" in a contest to produce the best results:

A group of researchers set out to answer this question, and published their results in Science last week. To tackle the issue, the researchers set up a computer-based tournament based on Robert Axelrod's 'Prisoner's Dilemma' competitions in the late 1970s. In this type of tournament, entrants submit computerized strategies that compete against each other in a virtual world. Individuals, or "agents," with the most successful strategies survive and reproduce, while less successful strategies die out.

In each round of the social learning tournament, automated agents could choose from 100 behaviors, each of which returned a certain payoff. The payoffs changed over the course of the tournament, simulating changing environmental conditions that might render a behavior more or less useful. In any round, agents could make one of three moves: use a behavior they already knew (Exploit), use asocial learning to test a new behavior by trial-and-error (Innovate), or learn socially by watching a behavior that another agent was performing in that round (Observe). Out of the three possible moves, only Exploit resulted in a payoff; the two learning moves would only return information about how profitable the behavior was in the current environmental conditions. Social learning was especially costly; if Observe was played when no other agent was performing a novel behavior, the agent learned nothing.

The results, however, showed that the runaway winners of the contest were those that used "social learning" the most. In other words, they were the ones who took what, on the face of things, appeared to be the most "costly" move -- and focused on what was working best for others and then using it successfully themselves. In other words, yet again, we see that the strategies that make the most sense for the greatest output tend to be those where participants in a market have the ability to copy others. Now, this upsets those who may have come up with the results first, but as other studies have shown, it's rarely the exclusivity of patents that leads to that invention in the first place. So if you don't need exclusivity to invent, and a more open solution of copying leads to greater overall output and social benefit... what, exactly, is the reason for creating these kinds of monopolies anyway?

Filed Under: imitation, innovation

Who Cheers For C When It Wins A Popularity Contest?

from the we've-got-spirit,-how-'bout-you? dept

While there are a variety of ways to judge the popularity of various programming languages, the TIOBE Programming Community index has ranked C as the most popular language for the first time in 4 years. At the top spot, C has no official corporate sponsor behind it -- but it's still available for nearly every platform imaginable. And arguably, it has always been the most popular programming language if all of its variants (C, C++, C#, Objective C) are counted together.

Interestingly, though, other notable languages that have risen in popularity have done so under the direction of large corporations. For example, Apple is the main cheerleader behind the recent popularity of Objective C -- especially as this superset of C is now one of only 4 languages approved for coding iPhone/iPod/iPad apps. Also, Google's Go language has been getting noticed because it's a shiny new offering from everyone's favorite "do no evil" buddy -- and because it explicitly supports concurrent programming.

But with no major company pushing for C, it may be getting a bit long in the tooth as multi-core processors inspire programmers to increasingly use multi-threading techniques. C will certainly never die, but the last revision of C was adopted as an ANSI standard in March 2000. So it's been quite some time since C has been updated. In fact, the ISO's proposals for C changes also appear to be pretty conservative -- which is a good thing for stability and eliminating any confusion over what is supported. But will C be able to evolve and stay relevant?

Filed Under: c, programming

Memristor Technology Beginning To Mature

from the still-just-a-toddler dept

A few decades ago, the memristor was a theoretical circuit element conceived by Leon Chua, a professor at the University of California at Berkeley. The "memory resistor" retains its memory even without power, and it joins the the resistor, capacitor and inductor as the fourth fundamental circuit element. Memristor technology became a reality in 2008 with a lab prototype that proved the existence of this basic electrical circuit. Following that breakthrough, a flexible memristor was developed in 2009, but despite the apparent progress in research labs, these devices are not exactly ready yet for commercial production.

Still, the potential for memristor technology is remarkable. Memristors promise to create a new type of memory that can mimic brain synapses in a way that could improve machine pattern recognition. More practically, memristors can be used in storage devices that would allow computers to boot up immediately -- since memristor-based memory doesn't have the same power or mechanical restrictions as conventional memory technologies. Recently, memristors also have been used to perform logic functions -- which could bring memory and logic functions closer together for more efficient computation. Given that a significant amount of energy is usually devoted to shuttling data around from storage devices to processors, memristor-based processors could also store data and reduce data retrieval inefficiencies.

The hurdles remaining for memristors lie in manufacturing challenges for introducing new materials into existing lithographic processes. Additionally, there are cost issues, scale up issues and questions about the reliability of new devices based on untested technology. Very optimistic labs are predicting commercial non-volatile memory based on memristors starting in 2013. However, given that flash memory was invented in the 1980s, and SSD storage has only recently become somewhat price competitive to hard drives, it may take a couple decades for memristors to really gain traction.

Filed Under: memory, storage

Can We Come Up With A Better Way To Measure Innovation?

from the the-market? dept

For quite some time, we've pointed out that patents are not a proxy for how much innovation a company does. In fact, research has proven this pretty conclusively. And yet, because they're easily countable, the press and politicians love to use the number of patents as a proxy for how much innovation is happening. That leads to silly articles from folks who should know better, making statements like "Clearly, the global recession seriously hampered innovation in the United States" because fewer patent applications have been filed.

But, that leads to a separate issue. If you aren't using patents as a proxy, then how do you measure innovation? Tim O'Reilly is asking for suggestions on a measure innovation metric:

How might we construct a metric that would reflect the transformative power of the web (no patents), Google (nowhere near as many as their innovations), Facebook (ditto), Amazon (ditto, despite the 1-click flap), Craigslist, Wikipedia, not to mention free software such as Linux, Apache, MySQL and friends, as well the upwelling of innovation in media, maker culture, robotics... you name it: all the areas where small companies create new value and don't have time, money or inclination to divert effort from innovation to patents?

The problem is that I'm not sure there is any single metric that really works here -- especially when it comes to disruptive innovations. You could go with revenue, but one of the features of truly disruptive innovations is that they sometimes shrink direct markets (while greatly increasing the size of indirect markets). So that might not be very useful either. You could go with user adoption -- but that may be fleeting or possibly gamable.

Even with older successful technologies, I'm not sure these kinds of metrics would most accurately highlight how much their innovation meant. Telcos made lots of money, but much of the real innovative value of the telephone was what other businesses they eventually enabled. How do you calculate that?

The only real metric I can think of -- though I'm not sure how one would measure this accurately -- is how much you would have to pay customers to get them to stop using a certain innovation. If you went around and surveyed people, and figured out how much it would cost to get them to, say, stop using search engines or email or mobile phones or automobiles, you might be able to get a sense of the "value" of certain innovations. From there, as a baseline, you could potentially monitor the delta over time. Thus, as the iPod grows in "value," the value of a portable CD player would decrease. As mobile phone cameras got better, the value of portable cameras would decrease, etc. It would be a lot of work, but could give you a much better general sense of innovation and how it changes over time than any patents.

Filed Under: innovation, measurement, patents, tim o'reilly

Why Real Programmers Don't Take The USPTO Seriously: Doubly-Linked List Patented

from the oh-come-on dept

It's pretty difficult to find software engineers who take the patent system seriously. There are a few, but it's still pretty difficult. For the most part, they recognize that code is just a tool: you can make it do all sorts of things, given enough time and resources, but that doesn't mean that doing any particular thing in code is an "invention" that no one else should be able to do. And then, sometimes, they discover that something pretty basic and old has suddenly been given a patent. Brad Feld discusses his discovery that doubly linked lists were apparently patented in 2006 (patent number 7,028,023):

The prior art was extremely thin, only went back to 1995, and didn't mention that entire computer languages have been created around the list as a core data structure.  One of my first Pascal programming exercises in high school (in 1981 -- on an Apple II using USDC Pascal) was to write a series of operations on lists, including both linked and doubly-linked lists (I always thought it was funny they were called "doubly-linked" instead of "double-linked" lists.)  Anyone who ever graduated from MIT and took 6.001 learned to love all varieties of the linked list, including the doubly-linked one.  That was 1984 for me by the way.

Ironically, Wikipedia had great entries -- with source code no less -- about both linked lists and doubly-linked lists.  The linked list history goes back to 2001, well before the patent was filed.

Another day, another reason to question why software is patentable at all -- and to question who approves these kinds of patents.

Filed Under: doubly-linked lists, patents, prior art, uspto