Slashdot Mirror
Computing With Molecules
ruppel writes: "Scientific American has an interesting article on molecular computing. The article is quite extensive, covering several technological issues and visions for the future. It also lists references for further reading and some interesting links.
" The article is a great technical overview of what's actually going in nano/molecular/x computing.
probably the best solution to the heat dissipation problem is reversible computing. I looked on google and found some links:
- Zyvex has a bit on the subject
- a group who say they've made a reversible-architecture CPU
- Mike's BibTeX file. I have no idea who the guy is, but there are some papers on the concept.
quantum computing is very good for many tasks (bye-bye Mr. NP-complete!), but not necissarily as good for others. however, my guess is that nanotech will allow quantum computing to become practicalLea
it's a good idea for some applications, especially for self-replicating types, but there's no real way to stream instructions to several billion running around in a spacesuit (for example) without some getting a huge lag. this /is/ one of those things on my todo list, but I'm a little busy :)
:) ). I'd suggest taking a look at some of the more popular works of Drexler (mostly) and Merkle. there are all sorts of reasons to want nanotech in general, nanorobotics more in specific, and nanocomputing most specifically -- and not only becasue we'll need it to run all the other stuff.
and in any case, you need logic circuits to run this thing, no matter what, and there aren't any other viable options other than using nanocomputing. mechanical, electrical, whatever. you have to have something on the other end, or else the instructions aren't going to help at all.
another problem that people are beginning to run into is that nanotech won't interface with what we have now very smoothly. there's going to have to be a huge changeover from one type of tech to another. it's not quite as bad as quantum, but think that sort of a change.
in any case, there is a lot of promise to nanocomputing in just about every application, especially embedded or portable stuff (can you IMAGINE the mp3 players?
Lea
very close to the stuff I'm doing right now, though I'm just doing it because it's just sitting in the way of waht I really want to do, really :)
if you remember the "nasa" snakebot article, that robot was actually a copy of PARC's PolyBot. They have another robot (not completely built yet) called Proteo which is exactly the embodiment of something like this.
Lea
oddly enough, my nanotech research could use some of the molecular computers that I've been designing (that's just a sideline)
/really/ hard part. I'm thinking pretty pictures :)
what I really need to simulate are emergent behavior algorithms -- then I need to explain them to mechanical engineers, which is the
Lea
You can have new models of computation without busting your noggin against exotic new physics. One easy way to do that is with new algorithms. Public-key cryptography opened up all kinds of interesting opportunities, running on tedious conventional hardware.
A couple years ago, I heard a talk at MIT about amorphous computing. It is basically a way of thinking about algorithms and communication so that we can successfully program low-reliability hardware (ordinary lithography/silicon stuff) to get reliably high performance. It's approximately the art of coordinating behavior in the presence of noise and unpredictability.
As a side benefit, this work is applicable to a lot of different scenarios for nanocomputing. They assume systems where processors or the communication pathways between them may be unreliable, or where there isn't a regular geometry, so this is the kind of thinking you'd need to program plaque-cleaning bots wandering around your arteries.
WWJD for a Klondike Bar?
Actually in AI and machine learning, most algorithms are basically search algorithms that look for a particular function/hypothesis. The problem is that usually the problem they are trying to solve is NP complete, i.e. the search space of possible hypotheses is exponential. To get around this, learning algorithms make horrible (and I mean ridiuclously bad and wrong) assumptions and simplifications. An example would be the naive Bayes classifier which is one of the best learning algorithms and yet the assumption it is based on is just plain wrong, but it makes computation faster. Also neural networks are usually made only 2 or 3 layer deep, because of speed considerations. Having more layers would improve the "intelligence" of the algorithm a lot, but is computationally intractable. So, speed is currently a very important issue, while Godel's theorem is almost irrelevant (in practice, philosophy put aside).
Diamond is not a molecule -- it's chemical formula is just C i.e. one carbon atom. It is a crystal with its crystal lattice organized in such a way that it makes it really really hard. You are right about big single molecules though, in plastic production plants, when something goes wrong all the liquid in the "reactor" polymerizes and you get one big molecule meters across. Those are extremely hard compared to normal plastics and extremely tough to break up.
Building systems is where the real challenge lies. It the difference between having a transistor or simple logic gate and VLSI circuits. There are heat dissipation problems that were not even mentioned in the article.
transistors work well for what they do. We need new computing models (i.e. Quantum Computing) not just smaller/faster version of what we have.
The Economics of Website Security
MIT's Technology Review this time around is a special issue dedicated to alternative/future computing methods. One of the sections is dedicated to molecular computing as well on sections discussing possible uses of DNA and other technologies. Very interesting and accesible to those without any particular specialized knowledge
self-replication and self-assembly mean factories turn into tanks, without spewing toxic chemicals all over the place. we would probably almost entirely stop using roads for shipping (and transportation in cities) in favor of extremely fast underground subways. we can smear the roads with an extremely tough substance which essentially acts as a solar panel that you can drive on and lasts for quite a long time. it's a very good way to get power -- you take otherwise useless radiating heat from the roads outside, and you release it out your roof, and on the way it's done a little work. toxic waste? that's one of the easiest of all to take care of. think of the bacteria that scientists are developing to "eat" oil slicks. it's more than possible to break down, molecule by molecule, entire toxic waste dumps into basically whatever you would like.
disadvantages: grey goo. if something eats up the entire earth, the environment will go, along with everything else. there are quite a lot of people worrying about this -- we anticipated it, so it's likely we can take care of the risk (through blue goo or similar)
there are other advantages: perfect recycling at a molecular level, basically an end to cancer and many other lethal and debilitating diseases, a chance to explore our galaxy... there are disadvantages as well -- but the environmental condition is not likely to be one of them.
Lea
I need to get on the bandwagon. Everyone is computing with molecules and here I am like a dolt, using pure energy.
--
Have Exchange users? Want to run Linux? Can't afford OpenMail?
Linux MAPI Server!
http://www.openone.com/software/MailOne/
(Exchange Migration HOWTO coming soon)
Although the advances were encouraging, the challenges remaining are enormous. Creating individual devices is an essential first step. But before we can build complete, useful circuits we must find a way to secure many millions, if not billions, of molecular devices of various types against some kind of immobile surface and to link them in any manner and into whatever patterns our circuit diagrams dictate. The technology is still too young to say for sure whether this monumental challenge will ever be surmounted.
They've shown that it's possible to build molecular-level electronic components like an AND gate and a memory analogue, but it'll be connected them together to form circuits that'll be the real challenge.
Think of how many transistors comprise even the most simplest of processors nowadays, and the technology it takes to fabricate them. And then think of doing the same, but with individual components and connections consisting of only a few molecules. There's going to have to be some real advances made in the ability to manipulate matter on this scale before this sort of thing becomes feasible on a large scale where economies of scale can apply.
But at least the proof of concept is there, and work will advance quickly with the threat of the end of Moore's Law approaching. And yet again, so much for all of the "end of computing" doomsayers :)
Don't get me wrong -- I've get nothing against faster processors. But my computer is already fast enough to play movies and do 99% of anything that I really want it to. If I'm not doing numerical integration, do I really need it to get any faster?
You don't and I don't (I was running a P133 until it broke) but there are a lot of requirements out there for fast computers. Look at this story for an example of a problem domain where superfast computers are required. There are a huge number of simulation tasks out there which can always use more power in order to use better models. And that can lead to any number of new technologies for us.
There'll always be a need for better computers. After all, since the smallest computer required to simulate the Universe is the Universe itself, we can always build a better machine to better simulate things.
Molecular computing is a similar phenomenon. At this point it's not really feasible, but who knows what it could turn into? Once computers reach the microscopic level, then we can begin to see some cool things happening. It will be like the transition from vacuum tubes all over again!
Got Rhinos?
If you guys would just leave the computer room for a while, you might get some sense of proportion about what you're destroying. I happen to own a small ngwa in the limitless plains of Tanzania, bought with my share of the fees on a biggish corporate real estate settlement. It brings tears to my hard face to think of the despoliation that will be wreaked somewhere just as lovely, simply in order to produce more toxic shit so that somebody can play Quake a little bit faster.
-- the most controversial site on the Web