Low tech pure optical computers?

[Arugela]

I keep seeing things on how they need to find ways to make logic gates. Why don't they use low tech things and do it with light itself. Why not use color or EM bandwidths and mirrors or other things that effect the light and make it process information. Why do you need something external?! On top of it why not use ambient light to power it? Can we not make mirrors that are pure optical. As in there is no power till you hit the monitor. If not the monitor itself is also powered similarly. We could get different display types that fit this best. It seems like they are trying to make optical computers based on electronic computers. It's a whole new material and medium. It should be done from the ground up using only thing from the related fields already existing from scratch.

For gods sake, you can amplify light with a mirror... You could even have absolutely no electricity involved!!!(Besides a light bulb at night to power it.)

When you change medium you should completely rework based on the new materials. Not translate things. That is the stupidest way possible to do it!

They should be investing in small or cheap ways to make mirrors in an accurate or other needed manner.

There have got to be a million ways to use light to process information. Probably way beyond what we are thinking. Any quality of the light. Anything quality of anything attached to it. All methods of processing. Simple and redundant is the best option with light potentially. Nobody should be remotely referenceing terms from modern electronic computers when working on it. Or they are not thinking outside the box enough to fully utilize it or optimize it.

And from that perspective we only need a CPU and HDD equivilant to make an optical computer. And those can both be done on a path. You do not need a single device. You can have each of those things functions broken down into infinite pieces and use. I mean you could easily make a unary computer for each program. or for each function along the way. And that is still way too close to reference electronic computers and fundementally still narrowing the view too much.

Can the light itself be used to process? Even without mirrors and whatnot? What about instead of actively effecting the light you passively detect it in way that process info. Are there any natural qualities of light predictable enough to utilize.

Edited October 7, 2019 by Arugela

[kerbiloid]

https://en.wikipedia.org/wiki/Opto-isolator

[Arugela]

I'm talking about pure optical computers.

[kerbiloid]

Probably because to manage the light rays one needs either electromechanical scheme, or a mechanical one.

Like they communicate in Going Postal mini-series or The Blue World by Jack Vance.

So, the light will be the slowest part of the logic.

[Arugela]

What if you could make light read light... What is the simplest way to use light to process. Even if it's not profound as far as computational abilities. Just literally the simplest. Can you even in a small way make light bounce off light?(or something similar.) Then instead of mirrors you use the light stream to do the work.

Edited October 7, 2019 by Arugela

[kerbiloid]

Just now, Arugela said:

What if you could make light read light...

Photonic beams almost don't interact.

[Arugela]

But if at any point they do it could be utilized. How little may not matter. There could be lots of uses. Including in complex optical computers. Error checking for instance in the stream. Checked along the path. Low tech or simpler for cost or ease of application. It doesn't even have to be powerful. It could power a wristwatch.

Edited October 7, 2019 by Arugela

[kerbiloid]

Just now, Arugela said:

But if at any point they do it could be utilized. How little may not matter.

They interact when they are really well powered. A transistor eats less.

[Arugela]

What about different types of interactions like light interacting with physical medium and used for things. WE don't need to process binary data with logic gates like we do now. That is limiting optics. Hell you could send light through a field of charged particles in the air, through water, through bubbles blown in chocolate milk.

All you need is one predictable point minimum. That could be generated or passive and simply understood(generated by something non man manipulated.). All are usable. Even simultaneously. If you know a starting and end state you can use anything predictable in between as a form of processing. Looking at it as light is already too narrow.

If light interacts at high energy I would assume it also interacts at all levels. It's likely just a matter of detection. High energy just probably creates an instant within a detectable parameter.

Simple Example:

Use mirrors and transparent materials. Catch ambient light and focus it. Use passive interactions in light as a driving force for processing information. Reading of info in a certain order could be processing from a programs perspective. Anything that produces the end results needed is code! You simply read predictable info from a light stream as it is naturally generated. Even independed of special mirrors as you could read the light path to find predictable info. The same could be done with passive light in the air. Even in all light conditions hypothetically as long as there is some light.

The transparent material would be for the purpose of getting rid of particles in the air for the sake of reading information. Or at least non predictable ones. Assuming that is needed.

Such a device could be both a powerful computer(or not so powerful) and something like a lazer razor! It could also detect air and apply custom hair cuts in various ways to your body hair using even a potentially different function(s) of the light.

The fun part would be a mechanical means of using the light it reads. Of course if you could utilize light for most of that...(maybe all if clever.) New types outputs could be used. If you can remove some you can hypothetically remove all.

Edited October 7, 2019 by Arugela

[Thor Wotansen]

Could you make light-based transistors that require little to no power to operate?  Maybe, but the bigger problem is memory.  What are you going to use for RAM and long term storage?  I don't know of any way to keep light just hanging around waiting for you to use it.  Sure, we could certainly have a read only memory that could be used to execute programs with a purely light based system, but you won't be writing anything to memory without a high power laser or an optical sensor that reads to a more conventional disk.

[Terwin]

If you look here: https://en.wikipedia.org/wiki/Optical_transistor  you will see that while there seem to be a number of possible techniques for an optical transistor, none of them seem to be further along than the proof-of concept stage.

Although as some of them require quantum wells or microkelvin temperatures , they generally do not sound either low-tech or low-energy.

(the lowest tech one, as far as I could tell, relies on transparency caused by heating, which is both slow and energy-intensive)

[Arugela]

2 hours ago, Thor Wotansen said:

Could you make light-based transistors that require little to no power to operate?  Maybe, but the bigger problem is memory.  What are you going to use for RAM and long term storage?  I don't know of any way to keep light just hanging around waiting for you to use it.  Sure, we could certainly have a read only memory that could be used to execute programs with a purely light based system, but you won't be writing anything to memory without a high power laser or an optical sensor that reads to a more conventional disk.

I'm under the assumption you only need storage and CPU. And those thing can be done as seperate new types of resources. You don't need ram specifically. I would start fresh and think it out from what is optimal and possible from lights perspective.

Maybe for HDD something like magnetic but effected by light. Save the state on contact just from contact and not electrical by nature. Then it can delay to save into. If it's delayed use multiple raid like redundancy to make up the time difference. I"m sure something more exotic could be done. Or something similar. Crystals? What saves state on contact with light. What about breaking down storage into different parts. what is storage. What could be done for everything else down the line? Could you start by keeping the program loaded via running in a stream constantly? Maybe normal electronic can be loaded?

Load the program from other means and keep a light stream with the entire program loaded and running and other things to modify it on a parallel path or similar. At that point constant light sources could be the ram.

Edited October 8, 2019 by Arugela

[Nuke]

with how they are able to grow laser diodes on a substrate its likely possible to be able to do the same with fully integrated optical components much the way a microprocessor is fabbed.

[wumpus]

18 hours ago, Arugela said:

Load the program from other means and keep a light stream with the entire program loaded and running and other things to modify it on a parallel path or similar. At that point constant light sources could be the ram.

In the dawn of computers, one common form of memory was a "delay line".  Some slow propagating signal was chosen store the data (such as waves of mercury) and then sent out and back.  This had two huge drawbacks: you only could access the data at specific times (and then typically had to resend the data), and you only had tiny amounts of data stored this way.  I'm fairly convinced that magnetic core memory was probably the biggest single thing that let computers progress to where they are now.

And while it has nothing to do with "pure optical", Intel made noises a few years ago about sending the most important/highest rate signals inside a computer (presumably meaning on the motherboard) through optical lines.  I don't think they are even close to commercializing it, and I'm not sure if that is a high priority anymore.

[Arugela]

Yea, but if the light is a stream and along the path is simply the modifications of it to start with it may naturally work for light. All you have to do it basically have a stream of light. Especially if you use the light in certain ways. The original stream could be the program. If it's stored in a wide array of light to start and different wavelengths are holding different code in a way you could simply take modifications of the original down the path and have the entire program in the original light or lights. Not sure where the problems start though or how much data could be held. I wonder how much data you could get out of the light stream. Color atm is about 16million at 24 bit... I would imagine there are ways to advance this. If you go beyond visible spectrum and more than just light as a quality how much data could be represented? Pulling out code could be interesting too. Would you mirror all aspects of the code into smaller chunks and then have static light and get the data later? Or would you modify the light constantly. I assume there are a lot of ways to process with it.

The fun would be trying to manipulate or access the light without slow downs. Mechanically would be interesting. Especially with no delay and powered by light itself and not electricity. Even passive light. If you can manage to amplify it under all conditions it would be cool.

Are there mechanical low tech means to manipulate or make certain light appear at certain periods to make a program run? You could do the equivalent of punch cards. But with light. Could you make this interactable? I would imagine you could hypothetically use physical interference of light as an activator. which would mean some sort of intractable monitor or some light based keyboard or controllers. That could be done in very primitive or very advance ways. It could be easy to expand beyond normal controls this way and have limitless keyboards fairly quickly. Unless cost is a factor. I assume radically different solutions are the starting point with light. You do not need to make the equivalent of binary processing at all.

A program could be a specific light source with a bunch of configurable nodes settings. Simply propagate into needed software and data size. It might make on demand 3d and other things easier also. Don't need to worry about ram and storage sizes if you directly configure into light. Couldn't you configure things physically to almost infinite data sizes?

Or for a more interesting method a specially designed mirror that takes light input and subtly bounces light between itself into a greater amount of needed data to get a program running. Even with special light spots being the source of code at different points. A 3d mirror is one way with many light sources between. Although I assume more subtle ways could be used. I would imagine a lot of different forms could come out that could be configured amongst themselves or interconnected. Although I imagine greed would kick in and it would all be used in the most restrictive and controlling way to get maximum cash flow.

Edited October 9, 2019 by Arugela

[wumpus]

Primitive ways won't work at all without external reaction (people reacting to "telegraphs" sending lights).  One thing you seem to be missing is that computers simply need an amplifier *somewhere* in the operations or the signal will be lost.  When nerve cells either fire or not depending on input, they can't be driven directly from input: they have to be able to create (or at least amplify from the input) the signal output.

Ultra-high-tech ways don't appear to promising (although I've heard of plenty of work there trying to get post-CMOS chips working).  One of the tricky bits is that I'm fairly sure you need branching instructions to be Turing complete.  Quantum "computers" might avoid this by essentially executing massive loops as single programs (and have some standard computer feed them each "loop"), which might be some way to have a "standard" computer do most of the work and let your "optical computer" do the "loops" all at once (of course this would require the same tech that builds silicon chips).  This "do one thing very wide" idea would at least avoid the need for an amplifier, or really just put the "amplifier" in the [non-optical] computer that is running the optical computer and creating all the light at once.

You're also missing a lot with the whole idea of a TV screen having a lot of data.  The HDMI cable that runs to the TV  also has all the data the TV can show and isn't exactly cutting edge on the electrical scene.  While light does have the capacity to harness a fantastic amount of data (see Claude Shannon's work on channel capacity  [1940s]), trying to modulate light at that level seems next to impossible (then again, wifi has been running at microwave frequencies for a decade, who knows what the real limit is.  Just expect to spend decades in infra red before reaching "optical").

Losing binary is a bigger problem than you might think.  Binary is only somewhat important (SSDs ditched that decades ago), but the Boolean (thanks again Shannon) underpinning allows much of what we think of computer operations to happen.  You would have to come up with some sort of similar operations that light can perform (perhaps some sort of associated memory (hash array) function, possibly with minor (presumably non-Boolean) operations to form some sort of map-reduce function (map-reduce basically powered google 20 years ago, but they've since moved on).

Throwing away 80 years of work on computing is a bit much.  I'd suggest learning the very foundations of computing (Shannon and Turing) to understand why things are made using CMOS electronics and exactly what problems you need to solve with light.

[Guest]

Well, have you devised how to build the basic "optical" logic gates?

Also, who is "they" (first post, references "them" a lot but doesn't say who it is)

[kerbiloid]

They call them "they", because that's it.

Spoiler

@Aperture Science gave up trying to understand a @kerbiloid's post, and changed the profile text.

 

Edited October 10, 2019 by kerbiloid