Have they made even a single working logic gate? The video only shows flipping bits back and forth by direct manipulation.
Edit: I am happy to report yes:
"Last, we explore the metastructure as simple mechanical logic gates. Figure 8 (C and D) demonstrates the achievement of both “OR” and “AND” logic gate operations by using independent bistability in local elements."
They do in the full paper (not open source & at this time no one's put it on archive):
"Mechanical logic gates
Last, we explore the metastructure as simple mechanical logic gates. Figure 8 (C and D) demonstrates the achievement of both “OR” and “AND” logic gate operations by using independent bistability in local elements. To facilitate the reading of output information (see more details in fig. S17A and the Supplementary Materials), we use a supported height-adjustable flat plate on the top to cover a small region of the platform. Its initial state is set as an output of “0.” When the plate is even and elevated, it outputs "1," otherwise "0" for the cases of either being tilted or lowered. The configurations of the top plate are determined by the pop-up ("1") or pop-down ("0") motions of three supports bonded to the bistable elements as inputs. A pyramid support denoted as P1 is placed in the center with two other neighboring supports surrounded, e.g., cuboids of S1 and S2 and pyramids of P2 and P3 for the OR and AND logic gate, respectively.
Figure 8C and fig. S17B show that when P1 is popped up and fixed, popping-up either S1 or S2 or combined as inputs leads to a stable and evenly elevated plate on the top as an output of "1" for an OR operation, because one point contact at P1 alongside one plane contact at S1 or S2 will render a stable and even surface. For the case of AND logic gate shown in Fig. 8D and fig. S17C, three pyramid that supports Pi are free to pop up or down, providing the point contacts to support the top plate. Only when the plate is supported by three pop-up point contacts, i.e., P1 = P2 = P3 = 1, it will generate a stable and evenly elevated plate as an output of "1" for an AND operation.
We note that most previous mechanical logic metastructures are limited to 1D and 2D structural forms (1–3, 11, 19, 20, 26, 29, 41). Our design extends the structural form of the mechanical binary logic computation to a 3D structural form. In Fig. 8 (D and E), we demonstrate the logic operation in only one zone. In particular, given the independent bistability of each local elements, such design principles can be readily applied to multiple zones for conducting a myriad of parallel mechanical binary operations on the same metastructure platform (see details in fig. S18). Moreover, by altering the structural components as schematically illustrated in fig. S19, we can also conduct “NOR” and “NAND” binary logic computations in our designed platform.
I believe a NAND gate is required as the base on which all other possible circuits can be built, so they need just to add an inverter for potential turing completeness.
It's not, we just use NAND everywhere because they're easier to make with transistors. You can get functional completeness with a NOR instead, or alternatively with some different combinations of other logical operators.
We even implement AND gates with NANDs in electronics (because they're way simpler), but we might not have to limit ourselves to a single base gate with mechanical computers.
> Where do you think the N in NOR and NAND come from?
That makes it sound like you could also do a NOT with XNOR, which is only the case if you can use a constant 0. But that would similarly also be the case for a XOR, but with the requirement of a 1.
“Second, this proof-of-concept work focused on binary computing functions with a cube being either pushed up or pushed down – it’s either a 1 or a 0. But we think there is potential here for more complex computing, with data being conveyed by how high a given cube has been pushed up. We’ve shown within this proof-of-concept system that cubes can have five or more different states. Theoretically, that means a given cube can convey not only a 1 or a 0, but also a 2, 3 or 4.”
Is this trying to straddle the line between analog and digital computing? Because it sounds like they are describing a crippled analog computer system.
Digital doesn't mean binary. A digital system must simply occupy a fixed number of levels, rather than the contiguous values of an analogue system. Binary systems, with 1 and 0, just happen to be one example of that.
There's plenty of non-binary schemes used in modern digital systems though. Modern NAND flash is one instance, for example QLC SSD drives store 16 distinct levels per storage cell (allowing each to encode the equivalent of 4-bits of data). Another example is 64-QAM, a modulation scheme used in a variety of places, including 802.11n Wi-Fi and Digital Terrestrial television (among others), which forms symbols out of two out of phase sinusoids, each of which can take up to 8 amplitude levels.
And even electronic computers haven't always been binary, one of the early Soviet computers was ternary, relying on 3 digits rather than the more familiar 2, to do all of its core computing functions.
Some existing digital storage uses multiple levels beyond 0 and 1 for improved density. For instance, the 8087 floating-point coprocessor used a ROM with four levels to store its microcode with two bits per transistor, as a regular ROM was too big for the die. Flash memory uses multi-level cells with up to 4 bits per cell.
> Mechanical computers are computers that operate using mechanical components rather than electronic ones.
For anyone who's excited about mechanical computers, perhaps it is worth reminding that an electron is about a thousand times lighter than a nucleon. Therefore, it's probably fair to say that mechanical computers will always be more energy consuming than electronic ones, because they fundamentally need to move atoms around to operate.
Taking this to its logical extreme, photonic computing should be significantly more efficient than electronic computing. Eventually.
Is that the end-game? Is there anything that would theoretically get closer to the Landauer limit than photonic computing? It’s way out of my element but I suppose this is a good venue to ask the question.
The big problem in photonic computing is actually making an optical transistor, i.e. a switch where the presence of photons coming from one source controls whether of photons coming from another source pass. This is harder than electrical transistors because photons are bosons and don't interact with each other, so even theoretically this is hard to imagine.
Papers that claim some progress pop up every once in a while but I haven't seen anything promising yet.
Yes, general photonic computing is mostly “theoretical” at the moment. Still, discussion of theory is important. I wish I could add more to your comment but I’m so far out of my depth that it would be simply misleading (blind leading the blind). I believe there’s theory saying it’s theoretically possible to create efficient photon<->matter interfaces which could achieve transistor-like behavior … but there’s too much I don’t understand to be able to evaluate whether there are inherent limitations which kill the practical application of the proposed theoretical mechanisms.
I think companies have come up with some practical applications of limited photonic “computing” at interface edges but I’ve heard that until we no longer need to convert photonics to electronics it won’t surpass electronics for general computing.
Possibly a strech, but transistors are basically current amplifiers, so their optical equivalent should be... lasers. Indeed lasers are optical amplifiers. Whether or not they can be turned into logic gates as transistors can, I don't know.
Maybe not more efficient, but maybe more resilient to electromagnetic storms, not prone to overheating (maybe), etc... Maybe it's about fitting constrained scenarios.
It seems like they may be prone to overheating in some fashion. All that electricity and motion has to cause some kind of thermal load. Or am I way off base?
Yes. Friction is usually the limiting factor in mechanical systems. It causes a lot of heat, noise, stress, and wear on all interacting parts. It requires all sorts of messy approaches to mitigate, such as lubricants and bearings. Electricity is basically magic by comparison.
Wouldn't that all depend on how much energy is used for computing, and how much for fetching and storing the bits involved? If the requirements involve slow computation with extremely long-term storage, perhaps mechanical computing can theoretically have an advantage.
Then again, Chuck Moore's GA144 shows there's still plenty of room when it comes to optimizing electron-based computing for those kind of extreme scenarios as well.
What if you made a really big circuit consisting of a battery, switch, lightbulb, and a wire that goes out 300k km on either side making a circuit that should take 1s at the speed of light to travel through. How long after closing the switch will it take for the light to go on?
“Few”, yes. But definitely some. I don’t think you can have propagation of EM wave through a conduit without at least pushing one electron into the conduit and removing one electron from the other side of the conduit.
I was pretty surprised about this since I had mistakenly believed that electrons had a velocity near the speed of light, which I think is only true in particle accelerators.
Indeed - I thought most college Physics 2 courses teach that electrons actually move quite slowly through conductors. It’s the “wave” which propagates near the speed of light, not the particles.
My mistake was being a biologist, and skipping or sleeping my way through the EE part of physics :) and then saying the wrong thing in front of some very smart people
I've had something similar- when I was deciding what grad school to go to, I was explaining how RNA enzymes work to some professor at UC Boulder, who ended up being Tom Cech (who won the Nobel for discovering RNA enzymes); he had to correct a lot of the details I messed up. I ended up going to UCSF and fortunately didn't try to explain prions to Stanley Prusiner.
In short, nearly everything I have learned is from saying dumb things in front of very smart people who instantly understood my misunderstanding and knew exactly how to explain it so I understood. That includes Sanjay Ghemawat and Jeff Dean telling me "your idea isn't so good, it's n-squared, here's a linear solution"
AlphaPhoenix did an amazing experiment to measure the speed of electricity FWIW. His other videos are incredible as well and explain EM physics in an absolutely outstanding way.
If I ever become a billionaire, I am going to have an entire big room in my house dedicated to pre-electric computers. It's amazing how much stuff got borderline-trivial once the transistor became ubiquitous, and stuff like The Writer Automaton has always been something that has utterly fascinated me.
If I ever happen upon a large fortune, I'd gladly give it to you if you used it for this and let me play with all the computers until I knew how they work. I studied mechanical engineering, and work in other fields now. Although I see and understand their importance to society, I am always fascinated by manipulation of mechanical principles to reach an objective.
Has anyone ever done a society re-bootstrapping from "mechanical computation" back to working chips? As in-a planetwide em-event knocks out all computation- how do we get factories like TSMC back on track? How do we keep food production going?
Could we recover from zero, using only this set of mechanical computers and basic instructions
I'm not convinced society would be able to recover after a collapse. We've extracted the easy to get resources already and what's left is extracted and refined with some really advanced technology. We can recycle a bunch of stuff, but that will only get you so far. On top of that, most of our knowledge is stored digitally, much of it in proprietary formats. If we don't somehow manage to recover it before the existing expertise dies out (assuming enough of it survives the initial catastrophe) we may not get it back for a very long time, if ever.
We would be at a huge disadvantage on the energy front. Most easily accessible fossil fuels are gone. Some countries have oil deposits that can be reasonably extracted now that the holes are made, but those places are few and far between.
On the other hand there would be vast resources of refined steel, aluminum and other metals. The average home contains materials that would make a monarch from 200 years ago jealous. Not to mention invaluable machinery like precision lathes. You just need to find a way to power them.
Without a readily available source for fertilizer and the supply chains necessary for modern agriculture we couldn't possibly feed more than a billion people or so. But whatever society rises from the ashes of that catastrophe could use the abundant building materials to harness water and wind energy and climb back up the technological ladder.
Computing would be pretty low priority though, first we would need to get farming back on track. Without modern farming you need most of the population to work in agriculture, preventing you from making any significant progress in other fields.
Wind. Wind, by itself, without storage, isn't great for our version of civilization. However, it'd totally be workable for 19th c, at least, if not early 20th c. If there were clever storage solutions, you could run our civilization, but at a lower per capita energy budget, to begin with.
Most of the important resources we have extracted are sitting around on the surface. When the steel rusts, it turns into some of the best iron ore. Most of the stuff we threw away ends up in landfills, mining those will get lots of resources and knowledge. That would be plenty for pre-industrial civilization.
The big problem is used the easy energy for industrial civilization. Solar mirrors and wind would be possible, but low density until more advanced. We assume that our energy-heavy industrial civilization is the only way, but it is possible that low-energy or low penetration industrial is possible. There is also possibility of biologically developed civilization.
Lots of current knowledge would be lost but there are tons of books from current and earlier eras. If those are preserved, there would be plenty of knowledge for early industrial civilization. In fact, the main problem would be finding anything or getting caught looking at past. One thing we could do today is make more durable books, and then reprint the important things like practical knowledge.
> Most of the stuff we threw away ends up in landfills, mining those will get lots of resources and knowledge.
I actually think this is something that’s inevitable for us today. The amount of high-value material that we’ve “discarded” by collecting it into one place and then ignoring it - the processes need to be developed, but at some point we’re going to recognize how much useable stuff we’ve just been piling up in the corner.
> I’m not convinced society would be able to recover after a collapse.
Like ever? Or do you have a time frame in mind? Our current state is evidence enough that people can get there eventually. This is all fun imaginary speculation, of course, but I’d wager that even if we lost all written/stored information and the scientists and engineers, just knowing what was possible puts the remaining people way ahead of where we were in the past. We didn’t know what was possible the first time through, didn’t know what to look for. Having memory of what existed and even a child’s understanding of how it worked, passed by word of mouth, would probably be enough to dramatically accelerate progress compared to it’s natural development.
Ever. Depends on the collapse, but one view is we'd never get back. The
easy sources of energy, ie coal and oil, have all been mined/drilled already, and you need those to jumpstart/bootstrap civilization to the point where you're able to produce enough food in order to have excess capacity of workers so you're able to get to renewable energy. Without those easy sources of energy, you don't get to have a post-collapse industrial revolution and you're stuck repurposing what's left over from the before times, and hoping it doesn't break because you can never make a fab to produce ICs to replace computers that break.
Generations after the collapse, the stories of what's possible will be viewed the same way we view stories of dragons from the middle ages. Fiction.
This seems to be imagining some other much bigger kind of collapse than what the thread started with. The top post proposed an EM event that knocks out computers, which is not something that destroys all books and mechanical machines, nor kills any people immediately.
I don’t necessarily buy the energy argument. Why would it have to be coal & oil primarily? Maybe it doesn’t. Coal and oil aren’t gone, but there’s also ample solar, wind and hydro to power a new society. Would losing computers actually cease coal & oil production completely? I kinda doubt that. I’m sure it would be a temporary setback and slow things down, but there was a lot of coal and oil production before computers.
The hypothetical question here seemed to already assume that food production and energy aren’t gone, it was just whether we can rebuild electrical compute without computers, based on knowledge of mechanical compute.
We can certainly imagine some epic worst-case scenario where no people with knowledge of any engineering survives, no books survive, and future humans have to start from absolute scratch. That seems far less likely than the probability that some of our knowledge carries. But even in the total doomsday scenario, what we have already is evidence that it worked the first time, and yes we can imagine hypotheticals that make it harder, but we already survived stories of dragons once, right? The default assumption kinda has to be that it might happen again given time.
When Cortez landed in Mexico and found the Mayan pyramids, they were ancient relics and the locals had no idea how they were created nor maintained. The knowledge was lost. Similar with the Egyptian ones. We know they could be built; we still don't know how.
Sure we do. Humans today could easily build a new pyramid if we chose to. You might be conflating the question of proving exactly what they did with the question of whether we could achieve a similar result today, those are two very different things and we’re discussing the latter. There is written evidence about the construction of the pyramids in Egypt from 4500 years ago, and anthropologists have have multiple plausible techniques with evidence (in part because we know there were multiple different construction techniques employed). If the knowledge was lost at some point, it’s not anymore.
Cortés and the conquistadors could have done it too, without knowing exactly how it was done before. Just having seen them, he’d know it could be done, and if he cared he could have figured it out. The Spaniards didn’t want to Mayan build pyramids, they were busy building castles and monuments - their own version of pyramids.
The technology to build large stone monuments has been repeatedly rediscovered. Once you have stone chisels, a quarry, and a log supply, plus enough food to maintain a population of people to work the stone and haul it, and a stable leadership, you can rebuild these in a few hundred years.
What do you mean, done a society? Like a simulation or a thought experiment (plenty, I’m sure, with “varying” levels of rigor). Actually run the experiment? I’m sure not.
Anyway it is generally impossible I’m pretty sure to do a “let’s build a society” experiment. Even if you try really hard, it always favors strategies that have a positive expected value but an unacceptably high chance of failure, right? Like you know the worst case if you actually fail is that you return to the real world and go to the hospital, so it is fine to take a risk that would give you like a 5% chance of getting an infection and dying. This has a 95% chance of working out but you’ll get a critical failure if you roll the dice over many generations.
Babbage went down the wrong rabbit hole. The technology of the 1860s was sufficient to make an automatic programmable computer. If you can make (or salvage) wire, and make simple sheet metal parts (brass will do) then you can make electromagnets, which means you can make relays. And there you go. No multimedia streaming, but automatic control and communication at a distance, yes.
BTW one of my favorite crazy ideas is that by the times of the Middle Kingdom, ancient Egypt had all the material resources needed to build a phone system, or at least a telegraph system, very useful in a large country. Zinc and silver to create batteries. Plentiful copper, gold, and silver to create any kinds of wires, and techniques to finely process it. Some amounts of magnetic ferrous alloys from meteorites, and likely access to iron ores to produce more. Only very small amounts are needed for kernels of electromagnets and membranes. Paper and resin-based glues could be used to produce wire insulation. Very certainly they were able to work any available materials with good precision and sophistication.
What they lacked was a good theory required to connect the pieces into a working phone system, like that of late 19th century.
I’m actually curious about the linguistics of something like a telegraph for ancient Egypt - I’m not particularly familiar with the Egyptian writing system from that era, but a strongly pictographic language, and one in which they seemed to regularly intersperse actual pictures and do things like manipulate symbol size to convey information doesn’t seem to immediately suggest translation to an encoding like Morse like phonetic written languages do.
In other words, if the ancient Egyptians had found themselves with the technology to create something like a telegraph, I wonder what they would have done with it - what possibilities suggest themselves given the visual representation of the Egyptian language.
(I could actually see something like the Incan quipu being a much easier translation, if we’re talking premodern “written” languages)
Getting off topic here but, the Ancient Egyptian writing system was only marginally pictographic. The fully-developed system is essentially phonetic, using about ~40 symbols to represent sounds, and a couple hundred pictograms (often interchanged with the full spelling). Closest analogy is probably the Japanese writing system, but with many fewer kanji. Carved hieroglyphs were highly formalized and ceremonial in nature, but ultimately, the quail chick means "u" and a foot means "b". (We inherit the letter "b" from the shape of that foot hieroglyph.)
I highly recommend that you read the book “Hieroglyphs: A Very Short Introduction” which I expect you will enjoy as much as I did: https://academic.oup.com/book/470
In short: Hieroglyphics were phonetic, but they eluded translation for centuries because only a small number of people could read and write them, and (importantly) the directions that the pictographs faced determined the direction that you’d read in.
My favorite fact from this book is that the hieroglyphic word for “cat” is the combination of the sounds for “me” and “ew”
The Very Short Introduction series is fantastic - they really do a great job of distilling the core of a subject to give a lay person the conceptual framework to appreciate the topic. I’ve enjoyed every one I’ve read.
> My favorite fact from this book is that the hieroglyphic word for “cat” is the combination of the sounds for “me” and “ew”
The Information Super-Nile. I like it. Perhaps elements of language used in cognitive expressions would begin to mirror through metaphor geographical terms used for proximity of major Nile features to the capital. Messages would become rafts. System operational periods, seasons. Fog of war would become a stagnant pond, or stilled flow. Riparian plants, ever-present information service providers such as scribes and couriers. Riparian birds with fleeting habits, commercial traders waiting to pounce in to action based on on news from afar.
- Make a list of the most common (spoken) words
- Sort them by usage frequency (most used words on top)
- Associate a unique sequence of 0 and 1 (dot and dash)
- Use shorter sequences for common words, longer ones for the words below
- As you don't transit letters but words the transfer rate is much higher than in "modern" telegraph systems
- profit ?
The pyramids were original covered in polished white limestone casement, which was highly reflective and visible hundreds of miles away. I wonder if you could exploit this to send pictograms by selectively covering parts of the casement with dark covers.
> As in-a planetwide em-event knocks out all computation
This is an impossible scenario. There will always be working computers somewhere, either shielded, excluded from the disaster, or by luck. Now they may not be the easiest thing to get to, and if all you can find is an iPhone or some other trusted compute platform you're probably SOL.
A scenario where general-purpose "unlocked" computers become extremely rare and nearly every computing device is a non-user-programmable appliance designed not to function without remote servers and regularly cycled cryptographic keys seems possible, though not inevitable. It would be a very stupid kind of apocalypse to live through if a solar storm didn't destroy many actual computers but did enough damage to brick 99.999% of the computers that survived intact, and left civilization with a crippling bootstrapping problem of capital's own design.
>if all you can find is an iPhone or some other trusted compute platform you're probably SOL
This reason alone should be impetus, for national security reasons, to severely tax products containing universal machines the device owner does not have full control over (by additional purposeful actions beyond owners' simple ignorance of the technology). I say a 100% sales tax on the retail price is fair, all tax proceeds going to fund GNU-compatible competition to the likes of iPhone and Playstation (and your proprietary microwave, and car, and TV, and ...). A post collapse society having to deal with iPhones hopefully will adopt the [corporate] death penalty for proprietary shenanigans like this.
It is borderline treasonous how many people will die because your product's bootloader was locked! Unforeseen consequences, Gordon.
> and if all you can find is an iPhone or some other trusted compute platform you're probably SOL.
Good point. The computer age is already poised to become a historical dark age; increasing adoption of trusted computing is only going to make this more severe.
On the other hand, some of it is necessary; on the other, security is the sworn enemy of sugar, spice and everything nice.
CollapseOS [1] comes to mind, but that's more about bootstrapping simple chips for directing power as opposed to rebooting society, but one could argue that the two go hand in hand.
The real deal breaker is probably the part where it would confabulate details in processes where the details are important. I'm just imagining trying to learn farming techniques from an oracle that might instruct me to make fertilizer that sounded sensible but caused nitrogen burn.
The big problem is that modern chip manufacturing is impossible without the petrochemical industry, which itself needs the petrochemical industry to operate.
The modern industrialized economy is built on industrial chemical production which stems from oil. If we lose the ability to extract or distribute oil it's going to be hard to bootstrap society.
But if that happens we're all going to die from common infections and starvation before we worry about getting YouTube back online.
Most post-collapse writing I've seen presumes a very long tail of salvageable computation devices not to mention batteries, LEDs, motors, solar panels, plus tools like oscilloscopes and soldering stations, etc that would all assist in the re-bootstrapping effort.
Assuming a planetwide em event is a pretty major wrench in the works and definitely puts you back a lot further in terms of how to rebuild the technology ecosystem.
I love this book because it does so in such a playful and imaginative way that you might not realize you are learning exactly how a computer works. But you are..
>Has anyone ever done a society re-bootstrapping from "mechanical computation" back to working chips?
Thankfully not, since it hasn't been necessary yet... At any rate I would guess mental arithmetic would be much more practical than mechanical computers, and then we could probably skip straight to vacuum tubes and punch cards.
Reading this thread I feel like I'm alive in the Star Trek universe. yesterday an article about warp drive, now discussing rebooting society after a global event and optical computers... interesting
Computers are measured in MHz and GHz. How do you even get close to that using mechanical means?
Speed is also a core value proposition (to use contemporary parlance) of computers I.e computers can carry out calculations at a rate of MHz/Ghz. If you want to talk in any meaningful sense of “computation”, and the usefulness thereof, speed of computation is a key metric.
I reckon cooling would be an even bigger concern than it is currently.
For many years I've wanted to build a digital clock out of fluid gates. The medium would be clear water; the 7-segment display would consist of transparent tubes, into which stained oil is pushed to activate a segment. I think a gate would probably be some kind of vortex. I've never tried to build it, because
- I don't know anything about fluid mechanics
- The 'exhaust' is water, so I could only really operate this thing in the bath
Edit: I am happy to report yes:
"Last, we explore the metastructure as simple mechanical logic gates. Figure 8 (C and D) demonstrates the achievement of both “OR” and “AND” logic gate operations by using independent bistability in local elements."