Slashdot Mirror
The Nanomechanical Computer
eldavojohn writes "The BBC is reporting on a newly proposed type of nanomechanical computer that mimics J. H. Müller & Charles Babbage's work on mechanical computational devices — just on a much smaller scale. The paper is published today in the New Journal of Physics and cites three reasons to build a computer with nanomechanical transistors over bipolar-junction or field-effect transistors: '(i) mechanical elements are more robust to electromagnetic shocks than current dynamic random access memory based purely on complimentary metal oxide semiconductor technology, (ii) the power dissipated can be orders of magnitude below CMOS, and (iii) the operating temperature of such an NMC can be an order of magnitude above that of conventional CMOS.'"
Neal Stephenson wrote a book about this kind of tech back in 1995 or so, entitled "The Diamond Age" (or "A Young Girl's Primer or something like that). He envisions some pretty incredible stuff made out of this tech. Great book, lots of nerdy CS-type stuff in it. Go to the library and pick it up, very fun stuff. I think this one of his works is very underrated. If we can actually engineer stuff like this it would be impressive, indeed.
Intel transfer the difficult from Hadware to software, for get more power, programmer need more technology. -- chinaitn
Anybody remember the mechanical nanotech in the book "Cassini Division"? Though they did that as a protective sandbox measure against supersapient AI and uplifted humans...
It may be a more robust transistor when dealing with shocks or heat, but I wonder if the same claim can be made for material wear. They'd be some diamond-like carbon structure, sure, but do we really know how robust those would be under such conditions (billions of jitters/sec rubbing against other pieces)? It could cause part wear or moving of parts, I would think....
Still, good on 'em....
It's an interesting concept, but it is not in any way a "new" concept. It was, for example, explored in Drexlers "Engines of Creation" that is (in full) available online under http://www.e-drexler.com/ EOC was first published in 1986, so the idea is more than 20 years old.
1. Funny that you should mention that, given that Babbage, get this: never actually finished his machine, so he never actually delivered any value for the ample funding money he received. Other people get into the v2.0 syndrome after they completed one successful project. Babbage couldn't be even arsed to finish the first one (although, again, he did receive more than enough funding for it) before he started designing the second version. Then the third. Then the fourth. What is now known collectively as the Analytical Engine is actually a whole series of different machines: he could never be arsed to actually finish building one before he got distracted and started the next one. He kept at it until his death.
His machines _would_ have worked, if they had actually been completed. But he could never be arsed to. Whenever he got funding for one, he'd deliver exactly nothing for that money.
So, you know, maybe _that_ is why Babbage found the Englishmen somewhat reluctant to invest in his designs. Had he actually finished the Differential Engine, maybe people would have been more receptive to his next ideas. Maybe instead of bitching about his fellow Englishmen, it would have been easier to just deliver what he had promised. Just a thought.
And maybe we would have had programmable computers a lot earlier. But as it is, it took people like Konrad Zuse in Germany or Alan Turing and the other folks who built the Colossus computer in the UK, to get it started. Because they actually delivered something that worked. Bloody huge difference there.
2. The complaint about "slicing pineapple" is actually invalid too. Like many nerds today, Babbage was in it just for the fun of researching something new, and apparently thought that people should give him a lot of money just so he can have some nerdy fun.
Capitalism, even the 19'th century kind -- actually, _especially_ the 19'th century kind -- doesn't work that way. To get some funding, the question you must answer is, basically, "which of _my_ problems does this solve?" If that company is in the business of slicing pineapple, then, yes, a machine which peels potatoes is completely useless to them.
Governments too, while they do fund some fundamental research too, have a fiscal responsibility to the citizens they tax for that money. Especially in the 19'th century laissez-faire ideology, when the government was lean, mean, and barely funded to maintain the army. You can't seriously propose a tax hike just so Mr Babbage can play with something cool and high tech. So basically they too have to ask, "ok, so what do _I_ gain from this? Does it compute ballistics for our battleships? Total the census? Or what?"
You'll notice that the working examples that did get computing started, had a satisfactory answer to exactly that. The Colossus computer broke enemy codes for the UK army, and Zuse's machines did aerodynamics calculations for the German airforce. E.g., the Z2 was used to design glide bombs.
A polar bear is a cartesian bear after a coordinate transform.
I suspect that what is being thought of is actually relay technology - so let's call it a Turing/Von Neumann/Mauchly approach (Alan Turing was a pioneer of relay logic among his other achievements, and Von Neumann and Mauchly were both associated with relay calculators.) Although relay computers were effectively obsolete by the 70s, they persisted in industrial controls for longer because (a) they could be debugged by electricians and (b) they could tolerate levels of contamination that destroyed the electronics of the period. The last generation of ultra-clean sealed relays and mercury relays were extremely durable and reliable. They didn't have the power handling, size for size, of power transistors but they had less internal dissipation. As a simple example, I was designing equipment in the late 80s which had to switch a few watts at around 500VDC. Although there are transistors that can handle these voltages, the design of the switching circuit necessitated a hybrid device costing around $200. A suitable relay cost $10 and was immune from punchthrough.
I'm prepared to guess that there will be niche applications for these ideas - but as with the IC engine, the sheer accumulated R&D in electromechanical systems will mean that widespread adoption will never be economic. It's easier to duct cold air over an engine management system (as on my car, with a few $ of plastics) than it is to redesign the entire chain from logic to actuator to use a different technology. And the current density of flash memory suggests that the hill to be scaled by electromechanical memory is enormous. Back in the days when flash chips were 256 bytes and not too reliable, there might have been a chance. Now when 8GByte USB dongles are cheap and reliable, it will be a lot harder.
Pining for the fjords
The research presented in this article reminds me of a an abstract I read a while back about a team who developed an on chip vacuum tube micro-triode which used carbon nanotubes as field emitters. It might not be possible to build a computer out of them, but logic built from them would have some of the same advantages mentioned in TFA (high immunity to electromagnetic radiation, etc.)i el5/16/21940/01019936.pdf
Link (warning PDF) http://ieeexplore.ieee.org/Xplore/login.jsp?url=/
Wrong, actually. The machine that was built at the end of the 20'th century was built with the precision and tolerances of the 19'th century. Deliberately, to show that it was possible.
The precision argument is even more obviously false, when you look at the fact that very precise watches had existed for a long time. That's how they measured longitude before GPS. I use watches as an example, because they're cog-based machines too, and they required even higher precision. By the middle of the _18'th_ century (i.e., a century earlier than Babbage) even a pocket watch would already not deviate more than a minute per day, and the second hand gradually became common. (Previously they tended to have only hours and minutes hands.)
The first practical nautical clock, John Harrison's H4 was first used aboard the ship Deptford which set sail for Jamaica on 18 November 1761 and actually arrived there on 19 January 1762. That's two months and a day at sea. After all that time, it was only off by 5.1 seconds.
_That_ is the kind of accuracy that was already available a century before Babbage.
Babbage's design didn't even need that kind of accuracy, since it was essentially a digital device. All that mattered was how many teeth of the cog had moved, not also to do it within a very exact time interval. Half the sources of inaccuracy of a watch, didn't even apply there.
So, no, Babbage had no excuse. The production capabilities were there, the precision was enough, and standardization wasn't even necessary for a prototype. He just couldn't be arsed to actually deliver what he promised. Full stop.
A polar bear is a cartesian bear after a coordinate transform.
Look back further still, to Richard Feynman in 1959. Absolutely visionary stuff.
J
But aren't mechanical shocks more common for your typical computer? And won't these machines take far more damage from them than current solid-state ram?
I am trolling
Perhaps he's waiting for the $10,000, or perhaps he knows that theory is the important thing, and if it's viable, there will be many organisations vying for better and better implementations.
I for one don't consider science to be something that only people with money do. One has to wonder how many da Vinci's there would have been, if other people all had the resources he had. The renaissance itself shows that progress pops up everywhere, given resources. Doesn't mean the science wasn't there in the back of people's minds, waiting for them to get past the point of scraping together money for a loaf of bread, though.
I know this is /. and actually reading the article is unusual, but *I* did and came upon this:
A computing architecture made from nanomechanical transistors thus is competitive with 45 nm CMOS technology Note 2, while taking a step towards enabling reversible computing. (emphasis added)I would LOVE to see THAT happen!
<dream>Whenever a program crashes, just open the debugger, run it backwards until it gets "weird". Run it forwards and backwards again to isolate where it's broken. Of course, there are some problems with asynchronous signals (disk I/O, keyboard, mouse, etc.) but I can dream, can't I?</dream>
But seriously, could this just be something thrown in to help get more funding or is it an actual possibility?