Slashdot Mirror
Intel Predicts Ubiquitous, Almost-Zero-Energy Computing By 2020
MrSeb writes "Intel often uses the Intel Developer Forum (IDF) as a platform to discuss its long-term vision for computing as well as more practical business initiatives. This year, the company has discussed the shrinking energy cost of computation as well as a point when it believes the energy required for 'meaningful computing' will approach zero and become ubiquitous by the year 2020. The idea that we could push the energy cost of computing down to nearly immeasurable levels is exciting. It's the type of innovation that's needed to drive products like Google Glass or VR headsets like the Oculus Rift. Unfortunately, Intel's slide neatly sidesteps the greatest problems facing such innovations — the cost of computing already accounts for less than half the total energy expenditure of a smartphone or other handheld device. Yes, meaningful compute might approach zero energy — but touchscreens, displays, radios, speakers, cameras, audio processors, and other parts of the equation are all a long way away from being as advanced as Intel's semiconductor processes."
I can't wait to overclock those chips so high that I need liquid cooling! Sounds like a fun project.
"Almost" ?? As in "I almost saw one camel today"?
My Psion Series 3a computed "meaningfully" on a couple of AA batteries for days.
touchscreens, displays, radios, speakers, cameras, audio processors, and other parts of the equation are all a long way away from being as advanced as Intel's semiconductor processes
It may not be possible at all to lower the power consuption of certain devices below a certain absolute threshold. No matter how advanced, a WiFi device has got to consume at least the power needed to reach other devices. A backlit screen will use at the very least the power it emits in light, etc... It is not simply a matter of technological advances.
That said: amazing prospect. Hope it's not just bold claims no substance. It would really be fantastic.
I wouldn't say that Intel is sidestepping those problems because they're not THEIR problems to address.
Sounds like he's talking about "localizers". Probably similar stuff found in other books as well.
Yes, meaningful compute might approach zero energy — but touchscreens, displays, radios, speakers, cameras, audio processors, and other parts of the equation are all a long way away from being as advanced as Intel's semiconductor processes.
I think the author misunderstood what "ubiquitous" means. It means you can put serious computing power anywhere, including in places that don't have displays, cameras, etc. He's just thinking, "How far can they reduce the power use of my existing smartphone?" The real question is, "What completely new types of devices become practical when computing requires hardly any power at all?"
Also, the situation is better than he suggests. Bright, super high resolution LED or LCD displays take a lot of power, but eInk displays use hardly any power at all. That's why battery life is measured in hours for an iPad and in weeks for a Kindle. LTE radios use a lot of power, but 3G is fine for most applications, and even 2G is more than sufficient in many cases (not for web browsing, but for a device that just needs to exchange limited data with the outside world).
"I'm too busy to research this and form an educated opinion, but I do have time to tell everyone my uninformed opinion."
When people think of the limits of strong AI (if they do at all) they generally focus on how complex it must be to create.
Complexity, however, is not the limiting factor. It is the fact that existing computers, compared to the brain, are energy hogs of epic proportions. The brain's energy use is on the order of millions of times more efficient than even the most power stingy CPU. Even of we knew how to accomplish strong AI we couldn't power the computer capable of supporting it.
That is, however, unless Intel reaches it's goals.
If I cast a net 10 feet around me I have 9 devices with CPU's in them. Only two of them are full up computers. Another two are smartphone. So the other five are my amateur radio gear, cable box, Wii, Xbox and TV.
Interestingly the kitchen and dining room are the only areas with the fewest CPU's in it. The office has a half dozen ATMega's as Arduino platforms, a TI Chronos Watch, Stellaris Robot, MPS430, and the oddest of all the CPU in the Western Electric 1D2 pay phone I own.
Aren't there some fundamental physical limits on how low your energy usage can be for a given amount of information based on thermodynamics? Is it just the case that they're way, way less than what we're using now?
Regarding energy requirements for a display and touchscreen, those are both greatly reduced with glasses (which, owing to their small size, are also the devices for which low power consumption is most important). Glasses are much closer to the eye, and ideally can direct the light directly to the eye. Modern displays are designed for maximum angle of visibility - they spew light over 180 degrees, on purpose, so they can be viewed from almost any angle. They are inefficient by design. So glasses can use much, much less power for display because they can be optimized in a number of ways.
Obviously glasses cannot make use touchscreens either, but instead use voice input, accelerometers, etc, which are hardware that require very little power.
Better known as 318230.
Cutting down power consumption by some factor == "Almost zero"?
This reduction in power will easily be made up for by more and bigger applications.
I think this is a shot against ARM. If anybody should talk about low power computing it's them. ARM with new tech like 22nm 3D multi-gate transistors will be *really* low power. Not Haswell & co.
thegodmovie.com - watch it
Well you can always use Solar or one of these.
thegodmovie.com - watch it
There are plenty of alternatives to audio and video that use very little power. The kindles e-ink screen for example... there's progress in delivering audio directly to the skeletal structure of your head, therefor using far less power. Wifi, GPS and cellular signals are where the real problems lay.
Although for trivial operations, the energy requirements are absurdly tiny fractions of a joule, I might suggest that for modern complex computing that we perform today, those minimum energy requirements aren't going to be anywhere as near to zero as they expect.
The only way it will really "approach" zero, is if we start demanding less from computing devices. This may be happening in some areas already, but I wouldn't say it's a ubiquitous phenomenon.
File under 'M' for 'Manic ranting'
Computation is zero energy, for sufficient values of zero.
Since the energy inputs and outputs of digital computation are necessarily equal for all forms of computation not involving fission, fusion, zero-point energy, quantum teleportation, black holes and other such esoteric things in the computation process, yes, computation is zero energy. I would wager that the entire amount of energy created or destroyed in the process of computation by humans in all of history wouldn't amount to an entire Joule.
Computation does now require converting energy from electrical energy format to thermal energy format though. Since quite a lot of electrical energy is used to create thermal energy in the regular course of business naturally this means that peak computation electrical energy efficiency can be improved not just by increasing the computations per KWh but also by putting the computation in the place where you wanted the thermal energy anyway, or using the thermal energy once you have it for some other purpose. That way you get to use the same watt twice at no additional cost. Fortunately Intel is already on this one too.
Help stamp out iliturcy.
You don't really mean zero. There is a fundamental minimum amount of energy it takes to do a calculation. When Intel says "almost zero energy computing" how far over this limit are they actually talking about? 101% of the Landauer limit? 200%? 1000%?
Give me Classic Slashdot or give me death!
When John von Neumann and his colleagues announced the world's first general purpose programmable vacuum tube computer he was asked how many the world might need, and guessed about 24. He was right and he was wrong. 24 of those machines would have handled most of the serious number crunching then taking place. But the machines brought about a radical reduction in the cost of computing, and demand exploded.
this may cause an increase in energy used for computations. http://en.wikipedia.org/wiki/Jevons_paradox
Even better: Bio-luminiscent color e-ink, to remove the need to use as much power for a display that is yet still visible in the dark (though to be fair, there are some VERY energy-efficient booklights on the market which do the job well enough, even if a bit clunky). As for what kind of stimulus would be needed to trigger the bioluminescence, well, that's beyond my knowledge.
"Since quite a lot of electrical energy is used to create thermal energy in the regular course of business naturally this means that peak computation electrical energy efficiency can be improved not just by increasing the computations per KWh but also by putting the computation in the place where you wanted the thermal energy anyway, or using the thermal energy once you have it for some other purpose. That way you get to use the same watt twice at no additional cost."
A related idea I had: :-) You could have this or any other local industrial process be thermostat controlled (or predictively controlled, or timer controlled, or some combination), so if your house or facility needs more heat you run the process; and if your building is hot enough for your needs, you don't run it, thus using local industrial-like processes to regulate your homes climate. For processes that absorb heat you could do the inverse for air conditioning. You can do that with networked computers too, so if you need heat you do local computation for the network, if you don't need heat, you shut those processors down. Special processor units or industrial process units for various purposes could be designed to replace regular electric baseboard heaters or central furnaces. So, essentially, industry is running for no extra energy charge where people use electricity to heat, and it runs at a subsidy where people use currently cheaper ways to heat like oil or gas or wood. And sometimes you might want to produce stuff anyway, and so you would need to dump the waste heat or use it in some other way or store it in some thermal storage system like a water mass or sand mass or phase changing salts or other such system, with the stored heat being used as part of the thermoregulatory planning. Of course, if you insulated your home well, you might not need a furnace, so there are economic limits to this idea as people improve their infrastructure in other ways...)
http://hardware.slashdot.org/comments.pl?sid=2344998&cid=36859662
"(I'll give away an idea here as a patent-preventing disclosure that I've been hoarding.
This would totally change how agriculture was done. Instead of having lunar moonscapes like Iowa is part of the year, people would just produce their own agricultural liquids in neighborhood facilities or at home, using the local waste heat for other purposes as well. Most agricultural lands could be returned to wilderness. The total energy bill for a home might not go up very much using the above idea for thermostatic regulation. "
A 21st century issue: the irony of technologies of abundance in the hands of those still thinking in terms of scarcity.
Smaller. Smaller. Smaller.
Smart Dust, is what we're talking about - or at least the early iterations.
Weather sensors that flutter in the breeze and scavenge enough energy to remain active and transmitting at most times - and the swarm *always* transmits.
Flow control sensors that oil companies continually release into their pipelines to ensure that if there's a leak they'll know where it is in milliseconds - there's transmitting sensors outside the approved geometric area.
Microscopic "Sniffers" released into the wind, measuring and reporting the amounts of cannabis, cocaine, explosives, dangerous chemicals...
Sensors to detect fire. Sensors to find out if the gas tank in that burning building is leaking at all. Just point into an air current (strong fan or wind) and let them fly from your hands.
*True* microsatellites, measured in single-digit centimetres or even smaller. (I think there's a minimum useful size for a satellite, but it's greatly related to how many of them there are, also... You could have a continual swarm reaching through the low-energy planetary transfer network keeping in contact with quite small satellites in a mesh radio network).
Making Smart Dust *safe* might turn out to be more of a challenge, though... :-)
But "really-really-low-power computing"... Alongside bio/nano-tech convergence it's the beginning of the real microbots:
Invisible cameras, as a perfect 3D image of your head emerges from the small swarm of the tiniest insects you've seen hover around your head.
Robots navigating through your bloodstream, tiny as hell - yet you've somehow ended up with the processing equivalent of your (2012) mobile phone coursing through your veins and working on any health problems you have (mostly by monitoring, at least at first).
I'm sure you guys can come up with more stuff. Please reply if you've got any ideas :-)
I'm a dreamer, the world is my playpen. But hey, I'm a serious person, I can't dream all the time.
Never mind that they haven't done it yet, or that the idea may be just pie in the sky. They had the idea first, so they should patent it. That way, if by some miracle somebody does do it one day, they can sue the pants off of them. That seems to be the way things are done these days.
Why would the Oculus Rift need this type of low power consumption? They do know it gets "plugged in" right?
Then how come the newer iPhones have worse battery life than the old ones?
You wouldn't use a Turing machine to model the minimum energy need of calculations, as they are woefully inefficient; in the same way that you wouldn't model addition representing the naturals through the Successor function.
The Turing machine was (is) a reasonably good tool to create proofs for the existence (or nonexistence) of computations, as it provides a quite simple and general computation model, easy to work symbolically with. But near the minimum use of resources, it isn't.
Singularity: a belief in the "God" idea with the "demiurge" relation inverted.
A minimum amount of energy it would require to reliably change a single bit can be reasonably be derived from this. Although for trivial operations, the energy requirements are absurdly tiny fractions of a joule, I might suggest that for modern complex computing that we perform today, those minimum energy requirements aren't going to be anywhere as near to zero as they expect.
It's the Landauer's principle but it's an extremely low limit. To quote WP:
At 25C (room temperature, or 298.15 kelvins), the Landauer limit represents an energy of approximately 0.0178 eV, or 2.85 zJ. Theoretically, room-temperature computer memory operating at the Landauer limit could be changed at a rate of one billion bits per second with only 2.85 trillionths of a watt of power being expended in the memory media.
Live today, because you never know what tomorrow brings
I've been following Kurzweil for years, and I've familiar with reversible computing, but I've never heard of him commenting on that topic. His arguments mostly stem from exponential growth of power. Citation? I'd like to read his words.
It doesn't follow that computing becomes ubiquitous when the energy use goes to zero. What's missing is the *price* of the chips. It's not like it's going to be magically cheaper to make a low-power i7 than to make the current one. In fact, even with the current power usage, you could stick an ARM chip in almost anything and have it do useful calculations. I mean, it's not the *power* that prevents us from having internet-connected light switches, locks, smoke alarms etc. now. You could have "the internet of things" making a comeback, now that we have the address space. The "thing" is either going to be hooked up to Ethernet, in which case power is not an issue (PoE), or it's going to be limited by the power use of a radio.
About the only advantage that neurons have over transisters is energy consumption. Compare the amount of energy that a computer takes that is computationally equivelent to a human (20 watts versus millions of watts): http://www.scientificamerican.com/article.cfm?id=computers-vs-brains
One of the difficulties of implantable cyberware is supplying power to the devices. There are ways of harnessing energy from glucose within the body, but these are limited to something like microwatts. If "Almost Zero" means less than a few microwatts, then maybe we can start seeing some implanted computers for medical purposes.