Exponential Assembly Top Down Nano 66
NanotechNews.com writes: "The article describes a new milestone in the Top Down nanotechnology process: "Exponential assembly is a manufacturing architecture starting with a single tiny robotic arm on a surface. This first robotic arm makes a second robotic arm on a facing surface by picking up miniature parts ? carefully laid out in advance in exactly the right locations so the tiny robotic arm can find them ? and assembling them. This is an exponential growth rate, hence the name exponential assembly." Standard MEMS, the largest independent
high-volume manufacturer of Micro Electro Mechanical Systems and Zyvex created a partnership, the second article available here. This partnership could lead to a better assembling technology in MEMS and the Top-Down Nanotechnology and Nanolithography."
finally (Score:1)
dns style? (Score:3)
My .02,
MEMS - to - Nano (Score:2)
Hitchhiker's Guide was there first (Score:4)
Anybody else remember that scene in Mostly Harmless where Ford Prefect breaks down the door to the head editor's office?
There are little nanotech bots in the doorframe whose sole purpose in life is to wait until this happens. Then they crawl out of the frame, assemble each other into larger bots, rebuild the door, disassemble each other, crawl back into frame, and wait...
Anyhow, I know some people working with MEMS. Very cool stuff.
so what? (Score:1)
huh? (Score:1)
what you have is a top plate interacting with a bottom plate with exponential surface area! first the top plate touches 1, then 2, then 4. yay! dear sweet god, please, someone enlighten me!
My .02,
Re:huh? (Score:1)
Alrighty! (Score:1)
Re:They're gonna take over the world (Score:1)
We've got lots of those already. :-)/p>
Then what? (Score:1)
Richard 'God' Feynman (Score:3)
Programmable? (Score:1)
Perhaps you could grow them onto the underside of your boss's shoes and watch him slide around on a million tiny scuttling legs. A worthy use of the millions this must have cost.
How does this work? (Score:4)
Here is my question. Assuming the arms are stationary, it is reasonable to assume that they can only build an arm adjacent to itself (and if they move, moving would be a O(n) process).... This means that for any grid area n^2, there are(n+2)^2 adjacent squares.
Because of this fact, I don't see how these things can achieve any more than a O(n^2) growth rate, because the adjacent resources available to these bots would be O(n^2).
Anyone know how these buggers get around this limitation?
Intriguing concept. . . (Score:1)
However, it does seem limited to assembly of pre-fabricated parts. Still, it's another step on the road to genuine nanotechnology.
What would be nifty, would be to merge this technology with chemical assembly: i.e. the smallest manufacturable arm, with a range of active tips, which use enzymatic techniques or positional assembly to build even smaller. . .
Nice concept, but how do you .. (Score:1)
x....
xx...
xx...
xxx..
xxx..
xx...
And now where is the exponential growth ? Rather better to push them apart every step, but can you really do this if you want to use silicon/hard material based nanontech ?
x.......
xx......
x.x.....
xxxx....
x.x.x.x.
Re:Hitchhiker's Guide was there first (Score:2)
Please take note of this and correct it in all future nanotech related comments.
What I've always heard about nanotech is that the main idea is that the robots are self-replicating. The example that frequently comes up is one where you put a few robots into a vat of liquid raw materials (a.k.a. molten metal) and they start creating more robots. Eventually when there are enough nano-bots, all of them manufacture an automobile.
Re:How does this work? (Score:1)
Re:huh? (Score:1)
Re:How does this work? (Score:2)
"...Because of this fact, I don't see how these things can achieve any more than a O(n^2) growth rate..."
You assume that they use a 2-D surface. To remedy this problem, all they need to do is not use a 2-D surface or use a 2-D surface and when the arm is built, it get's transfered to a new location away from where it was built. An assembly line has always been an efficient way to manufacture products. Perhaps the robots would create assembly lines as they were created. Then in a 2-D space they would be able to obtain an exponential growth rate.
Re:How does this work? (Score:1)
Well, all that would do is make it a O(n^3) limitation, which would still be polynomial and not exponential growth....
Also, the act of transfering the arm is going to be at least a O(n) process which would also limit production. since there are three dimensions in which to move, this limitation would also be o(n^3) in nature. picture things moving away from each other at a constant rate but getting bigger at a rate of S=n^3. They would eventually overlap and consume each other.
Re:Hitchhiker's Guide was there first (Score:1)
Actually, the earth was never re-assembled. What actually happened was that the last few books of the "trilogy" took place in a different "parallel universe", where a couple of key decisions were made differently than in the first books: earth not demolished, and Trillian didn't go with Zaphod. No reconstruction, just different choices.
Re:How does this work? (Score:1)
Okay, I can see arguing the blender idea, beacuse one could say that it would be a limitation of the size of the blender that limted production, given enough random motion, but it seems that even this is sorta deceptive because of the lmiitations of 3-space.
The blender, as it mixes is after all going to increase in nanobot density which means there will be a amount of stuff around it, and mixing of an infinite blender doesn't really make much sense, because how would a particle get from one edge of an infinite blender to the other?
Why just scratch BACKS (Score:1)
This reminds me of an old short story (Score:2)
A guy built a robot that was supposed to build a single copy of itself at one tenth the size. He made an error in the program, and each robot built ten copies of itself at one tenth the size. The robots eventually got so small the would duplicate at a very high rate. The guys house was eventually destroyed, and the only thing that saved the day was a rainstorm that rusted the robots to death.
Anybody remember this one? Anybody got a link to it?
Re:dns style? (Score:1)
Re:Hitchhiker's Guide was there first (Score:1)
Umm... nope.
It starts in the mid-80's and continues from there.
It takes place off-planet, which is why the tech is so advanced.
-Ciaran
Please read the article (Score:1)
Reading the article [zyvex.com] you get the answer: The arms are put the surface of a plates. In front of this plate you put a plate without arms. Now the one plate puts arms on the surface of the other other plate. Afterwards you put each plate in front of a plate without arms. Repeating this proces will generate exponential growth. You would also get exponential growth if you had somehow put the arms in water and added new unnassembled arms, to maintain the same concentration of assembled arms.
Re:How does this work? (Score:1)
I'm not so hot on the math here, but could it still be exponential if the movement were taken into account as part of the mfg process (simultaneous or not)?
2nd motor: (Tmfg + Tmovement)
3 + 4th motor: (Tmfg + Tmovement)
5 - 8th motor: (Tmfg + Tmovement)
Re:How does this work? (Score:4)
However, your order estimates are incorrect.
This means that for any grid area n^2, there are(n+2)^2 adjacent squares.
Yes, and the rate of growth is determined by the difference between these two, which is O (n), not O(n^2).
Because of this fact, I don't see how these things can achieve any more than a O(n^2) growth rate, because the adjacent resources available to these bots would be O(n^2).
In fact, it is O (n). To easily visualize this, imagine the system in 1D for a moment. After the first unit assembles its nearest neighbors, each additional unit builds the next unit at the end of the line of units. This leads to a constant rate of growth. In 2D, the rate of growth is determined by the rate of change of the area, not length, which leads to O (n). In 3D, the rate of growth is determined by the rate of change of the volume, which leads to O (n^2). The result in each case is easy to visualize : it is limited by the boundary of the N-D volume the units have already filled.
All this said, I think this whole discussion doesn't emphasize that even an O (n) growth rate can be vastly enormous for large enough n. The main problem is that it appears much slower to start up than a truly exponential process, which could be realized for a longer duration if the newly built units were "mixed" randomly into the grid. This could be the case if each unit could be assigned to move some random distance under its own power, after each building cycle. Eventually, however, the exponential rate of growth will turn over when the system's capacity is reached.
Re:How does this work? (Score:1)
Yes! You are right! However I would like to save face and say that something that is O(n) is also O(n^2)... heh, technically. The boundary of a square is 1-D and a surface of a region in 3-space is 2-D... I see your point.. It's worse than I expected.
newly built units were "mixed" randomly into the grid. This could be the case if each unit could be assigned to move some random distance under its own power, after each building cycle. Eventually, however, the exponential rate of growth will turn over when the system's capacity is reached.
Wouldn't the fact that the systems had to move limit the process? After all, moving takes time, and the amount that they would have to move would be going up on the at least the same order as the size of the region.
It would be exponential (Score:1)
Since the arm is not building the replica on the plate it occupies itself, all that is required at the next growth step is for the plates to be moved relative to each other in such a way that each arm is facing an empty area of the opposite plate.
Re:How does this work? (Score:1)
Re:This reminds me of an old short story (Score:1)
Re:huh? (Score:1)
The manual process can make one arm per time-slice (however long that takes) using what looks to be a single, large, expensive device.
However, by building a single arm manually, and then moving the surface to another location you can breed more arms, at 2 per time-slice, then 4 per time-slice, etc.
Once you have 1024 [for example], you chop them up using another device, and place them on another surface in any arrangement you desire and get them to build things.
1024 manually takes 1024 time-slices, 1024 exponentially takes 11 time-slices.
Re:Then what? (Score:1)
Building them is the first stage, building millions of them is necessary, so building them quickly is essential.
Simplified, or what? (Score:1)
I'd like to see a macro-scale implementation of
this. Where are the servos that move the arms?
What about power distribution and controll
signaling? How are these attached to the arms,
and how do you make sure the power, signaling
and servo attachments don't get in the way?
If you've already manufactured all the parts and
laid them out in a perfect pattern on the two
surfaces, why not do all the assembly at that
stage? Surely this level of assembly is the
simples step in the manufacture process? So
simple that it's completely unnecessery.
Simon Hibbs
Re:Nice concept, but how do you .. (Score:1)
You do it like this - remember you can move the two facing plates as far as you like. On a 1-d surface, with space for 8 arms, you might get this (forgive awful ascii-art):
Step 1: Start just overlapping, and build:
_______1......._______1.......
Step 2: Slide half-way back, and build
___1.......___1...4...
Step 3: Slide half the remaining distance back, and build
_1...4..._1.7.4.6.
Step 4: Slide the one remaining slot back, and build
1.7.4.6.1D7B406A
95F3C8E2
Presto! 2^n robots in n steps, with no relocation required. Once you've made a line of robots like this, repeat by translating in the other dimension after each step, and voila!
In order to allow the final plane to do non-trivial stuff after its finished, you give each machine a unique address etched into the control electronics layer rather than the MEMS layer. But in order to allow the things to operate in lockstep, you can also have a "multicast" address (255.255.255.255?) which will address all devices on the chip. This would be useful even after manufacture as a master reset or as a way of shuffling finished components off the chip at the end.
Sean Ellis
This was proposed by Robert Heinlein in Waldo-1950 (Score:1)
Hey stupid moderator! (Score:1)
Therefore this is not offtopic.
Demoronize /. (Score:1)
Re:dns style? (Score:2)
Reading the article usually helps.
Re:They're gonna take over the world (Score:1)
Re:huh? (Score:1)
How do you "feed" a robotic arm? (Score:1)
You have to build also the "control" part in order to pilot the move of the arm, it can be as difficult to make as building the arm itself.
The next problem: how does the arm catch the needed molecules?
This may be easy if there is only one type of molecules: put the robot inside a solution of these molecules.
But if you need more than one type of molecules??
I can see an easy solution for the first arm, but for the second? How do you connect it to the energy source, to the molecules tank?
What I find strange is that nobody has designed a complete working auto-replication system with nano-bot.
Sure to a degree, the firt nano-bot design will be dependant of the way it has been built, but I think that trying to simulate as completely as how it COULD work would be an interesting baby step..
Grey goo takes over the world, film at 11 (Score:1)
Sorry.
bootstrap problem ahead! (Score:1)
Re:Alrighty! (Score:1)
This whole concept frightens me.
Nanobots!, a poem (Score:1)
I love you, Xev/Zev! (Score:1)
Aciel
aciel@speakeasy.net
Re:How does this work? (Score:1)
These nanobots are really tiny (hence the name) and you will need to make lots of them to do anything useful. If you are trying to make 10^15 of them, you dont want to have to make them one at a time, a thousand at a time or even a million at a time (all have a o(n) growth rate), that would take all too long. By having an exponential growth rate, they can focus on making larger plates to grow the nanobots on rather than making more nanobot producing factories.
Re:How does this work? (Score:1)
Your O(n^2)(which should be O(n) as pointed out in another reply) limitation is only on space as you state. However, the O(2^n) growth rate is a limitation on time. Thus, if you double the area of the plate, you can make roughly twice as many arms but with only one more time step, not twice as many time steps.
An interesting point, although I still don't exactly see how the things could grow exponentially at all without some kind of motion to move the "center" nanobots.
In a sense, its like having localized depletion of resources. The oldest bots would be surrounded by younger bots, and I don't see how this is avoidable, even given the complex geometry of these plates.
My contention remains that localized "starvation" of these bots would limit their growth rate to polynomial time. (Which aint bad)
gingerbread men (Score:2)
If you've already manufactured all the parts and laid them out in a perfect pattern on the two surfaces, why not do all the assembly at that stage?
As I understand it, you can mass-produce the components using micro- (and perhaps nano-)lithography like a cookie cutter, but what you end up with looks like a bunch of cookies on a cookie sheet. You need the robot arms to put the, er, cookies together.
It's not exponential (Score:1)
The answer, of course, is polynomial in T, and *not* exponential at all. For exponential growth the cells must move apart; in this algorithm, the early cells are quickly surrounded by other cells and can do no more work.
Re:How does this work? (Score:1)
A '1' can only build 2's(for efficiency sake), 2->3, so on, and a 'b' unit(second gen) can build a 'c', 'd', and 'e', with all resources used up. In other words, max useful assembly life of these things is t+3, because the one that built them is on one side.
Growth pattern
---4
--434
-43234
4321234
-43234
--434
---4
Lattice timing chart
_______|5F|__|3E|__
_______|4E|__|2D|__|4F|__
_|5E|4D|3C|2B|1A|2C|3D|4E|5F|_
____|5F|__|3D|__|3E|__|5G|
__________|4E|__
You get the idea, I hope. It isn't quite a Von Neumon(sp) machine, but getting there! The idea is to make two, then go out and explore, colonise and whatever. I think the nanos are on the right track, but How they do it is another thing! Why, at 100 microns, they could assemble circuitry! Imagine the non-lithography, no chemical advantage! neat!
I just can't wait for them to check microstructures for flaws, improving quality, etc.
Re:Demoronize /. (Score:1)
Re:Hitchhiker's Guide was there first (Score:1)
Re:Hey stupid moderator! (Score:1)
Re:I love you, Xev/Zev! (Score:1)
Re:It's not exponential (Score:1)
Re:Alrighty! (Score:1)
Consider that it takes an entire multinational industrial system to provide the power to a single tank and then ask yourself how much we need to worry about gray goo. In the la-la land of nanotechnology, little concerns like power source and heat dissipation are not interesting enough to consider, compared with fanciful speculations about Fantastic Voyage cell repair robots and smart mists.
Re:How does this work? (Score:1)
I dont see why you need to move the bots if you can move the plane. As it says in the article:
While each robotic arm would have only two degrees of freedom, the surface itself could be moved in X, Y and Z.
I know im assuming alot about the process here and I could be completely wrong on this, but if you move the planes in an organized way, it is plausible that you wont have to move the bots and that they wont crowd each other too much either. This is because each bot doesnt make a new bot on its own plane, it makes it on the opposite plane. Thus, bot1 on plane1 makes bot2 on plane2. Currently bot1 is still close to bot2 since the plates havent moved much. Next, big machine moves plate2 1 meter parallel to plate1. Now, bot1 is 1 meter away from bot2. Bot1 and bot2 have plenty of room to reproduce. Repeat till bots are too dense to reproduce. Clean up mess and start over.
Now if your plates are large enough, you wont have to worry about crowding for quite a while. The only problem would be in coming up with the scheme to move the plates and in making big enough plates. I guess you could call this "moving the center nanobots" even though you arent really touching them in any way, just moving the plates they are on.
It's quadratic or linear growth, isn't it (Score:1)
Re:How does this work? (Score:2)
I think that you could achieve exponential growth on a 2D surface however. Here's how...
1. Each existing robot builds an additional robot.
2. Each existing robot then relocates such that there is enough room to build an additional robot.
Technically, that is exponential growth (w.r.t generation or iteration). One can argue that the relocation step takes longer and longer each time, but it's still exponential growth.
It all comes down to whether your 'n' is an iterative or a temporal element.
I'm not trying to talk you down or say that your are wrong.
What about an assembly line that the robots moved down as they got built. The 'just-completed' robot at the end of the assembly line would then contribute to building the next robot. Since more and more robots were joining the assembly line, each robot could do less work. Since each robot was doing less work, the assembly line could move faster and faster, thus attaining an exponential growth rate.
I know, I know that assembly line would only approach the speed of light or some asymptote which was representitave of the robots' speed. But it's a cool idea.
exponential till you build something practicle (Score:1)
Sounds like pipe dream, why stop at nano machine robots, why not make a big robot that builds other robots... who hasnt thought of that before.
Heinlein already did this in "Waldo" (Score:1)
Re:How does this work? (Score:1)