How difficult would it be to build hardware that computes in tits {-1, 0, 1} instead of bits {0, 1}?

How difficult would it be to build hardware that computes in breasts {-1, 0, 1} instead of bits {0, 1}?

It's All Fucked Shirt $22.14

Ape Out Shirt $21.68

It's All Fucked Shirt $22.14

  1. 2 months ago
    Anonymous

    Use an fpga, store the values in two bits which gives you a hardware null for better or worse.
    You'll get the improved ternary performance, and because it's an fpga you can encode your registers and calc blocks to match the model structure. Might be difficult because of model sizes, so I guess you either get big ass fpgas or a lot of them

  2. 2 months ago
    Anonymous

    I computed all over your mom's breasts last night bucko. It was pretty easy.

  3. 2 months ago
    Anonymous

    if the soviets could do it, so can you

  4. 2 months ago
    Anonymous

    already exists doesn't help much. Also inb4 quantum bullshit 99% of the stuff online gets it wrong I want to know who spread that nonsense.

  5. 2 months ago
    Anonymous

    There's no need. Much like the paper does, you encode information in the extra .58 bits. IE devote some smaller portion of your binary systems for some compressed representation of signs.

  6. 2 months ago
    Anonymous

    Dunno, but you will have to rewrite everything from scratch and find a way to represent the third logical state you want as an electrical signal since it won't be just HIGH and LOW which is what everything is built upon.

  7. 2 months ago
    Anonymous

    >How difficult would it be to build hardware that computes in breasts

  8. 2 months ago
    Anonymous

    you need to build a logic gate that outputs 3 states first. how do you even invert -1?

    • 2 months ago
      Anonymous

      >how do you even invert -1?
      Make the voltage of -1 be below the voltage of 0.

      • 2 months ago
        Anonymous

        so now we have 4 states instead of {-1, 0, 1}

        • 2 months ago
          Anonymous

          What the frick are you talking about? You have three different voltage levels. Middle one is for 0.

          • 2 months ago
            Anonymous

            not(0) is 1
            not(-1) is ??

          • 2 months ago
            Anonymous

            >not(0) is 1
            >not
            You mean negation? -0 = 0.

          • 2 months ago
            Anonymous

            NOT -1 is 0 OR 1
            NOT 0 is 1 OR -1

            Get your head out of your ass.

    • 2 months ago
      Anonymous

      That doesn't matter.
      Bits have 2^2=4 possible logic gates with one input and one output and 2^4=16 possible logic gates with two input and one output.
      Some have cute names like "XOR" and "NOT", but they don't need to have names or even make sense to be useful.

      breasts have 3^3=27 possible logic gates with one input and one output and 3^9=19683 possible logic gates with two input and one output.
      More gates to choose from means more possibility to optimize, you don't have to restrict yourself to gates that make sense to you in the real world.

      • 2 months ago
        Anonymous

        did you read the OP? he wants hardware that computes in breasts.

        • 2 months ago
          Anonymous

          Gates aren't hardware?

          • 2 months ago
            Anonymous

            i just don't understand how to build logic gates that can react to 3 states. it is theoretically possible combining transistors with different threshold voltages, but you are still using binary logic.

          • 2 months ago
            Anonymous

            >i just don't understand how to build logic gates that can react to 3 states.
            You don't understand the concept of building circuits with input and output voltages that can alternate between 3 instead of 2 approximate voltage values?

          • 2 months ago
            Anonymous

            as i said in the rest of my post, i understand, but you are still using binary.
            it is hardware that computes in bits, but simulates breasts.

          • 2 months ago
            Anonymous

            >3-valued logic gate is 2-valued
            Almost every IQfy poster is a literal LLM.

          • 2 months ago
            Anonymous

            Learn some python and virtualize them make the virtual prototype of them and see how it will work.

            Virtual logic gates are super primitive to code and if you want something advanced just literally use the fricking switch.

          • 2 months ago
            Anonymous

            I thought OP's question was:
            >How much harder is "breasts" hardware compared to "bits" hardware?
            not:
            >Tell me in detail how to build a "breasts" computer.

    • 2 months ago
      Anonymous

      Hardware doesn't use "1" and "0", it uses "high" and "low".
      -0.1 and 0.1 are 0, but so is -9000.
      This lets it use brute force to set something to 0 or 1: just push it up or down hard.

      If you have 3 values setting the middle value requires a lot more delicacy: can't push up or down to hard or you end up with a -1 or +1.
      So it's much harder than binary and almost never worth it.

      That doesn't matter.
      Bits have 2^2=4 possible logic gates with one input and one output and 2^4=16 possible logic gates with two input and one output.
      Some have cute names like "XOR" and "NOT", but they don't need to have names or even make sense to be useful.

      breasts have 3^3=27 possible logic gates with one input and one output and 3^9=19683 possible logic gates with two input and one output.
      More gates to choose from means more possibility to optimize, you don't have to restrict yourself to gates that make sense to you in the real world.

      higher 1
      no change 0
      lower -1

  9. 2 months ago
    Anonymous

    The easiest solution with current machines is to just use 2 bits per tit. There's a bit of a waste since you're using a 4 state system to model 3 states, but what can you do?

  10. 2 months ago
    Anonymous

    Hardware doesn't use "1" and "0", it uses "high" and "low".
    -0.1 and 0.1 are 0, but so is -9000.
    This lets it use brute force to set something to 0 or 1: just push it up or down hard.

    If you have 3 values setting the middle value requires a lot more delicacy: can't push up or down to hard or you end up with a -1 or +1.
    So it's much harder than binary and almost never worth it.

    • 2 months ago
      Anonymous

      >low is ~0 V
      >high is ~X V
      This is easy.

      >low is ~0 V
      >mid is ~X V
      >high is ~2X V
      This is somehow difficult?

      • 2 months ago
        Anonymous

        >low is -infinity to 0.5
        >high is 0.5 to positive infinity
        This is easy: just aim for +1000 and -1000 instead and you can't really miss.

        >low is -infinity to 0.25
        >mid is 0.25 to 0.75
        >high is 0.75 to positive infinity
        Hitting that mid reliably is difficult.

        • 2 months ago
          Anonymous

          > just aim for +1000 and -1000
          I have an arduino here that hits 0 and 5 pretty consistently. Why shouldn't it be able to hit 10 as well?

          • 2 months ago
            Anonymous

            It's like playing darts and trying to hit either the left or right side of the board.
            But hitting the wall is fine too: left wall =- left, right wall = right.
            You should be able to do it even if you're drunk and have no talent.

            Then you have a new system where you have to hit the middle 33% of the board too.
            Now you actually have to aim.

          • 2 months ago
            Anonymous

            >It's like playing darts and trying to hit either the left or right side of the board.
            Do you understand what "5 volts" means?

          • 2 months ago
            Anonymous

            In binary?
            3V = 5V
            7V = 5V
            999999V = 5V

          • 2 months ago
            Anonymous

            Do you suffer from some kind of mental moronation? I have an arduino here. It delivers very close to 5V from its GPIO pins. Actual 5V. If that doesn't require any alien technology, why would also delivering 10V require it?

        • 2 months ago
          Anonymous

          >Hitting that mid reliably is difficult.
          More of a problem is you need to leave longer for the voltage to stabilize, and can't interpret from observations of current (you don't know if you're going form low to mid or high). That will reduce the safe clock speed.

      • 2 months ago
        Anonymous

        >This is somehow difficult?
        it's not difficult, it's inefficient

  11. 2 months ago
    Anonymous

    You can but each component would need to be custom since it would be incompatible with all the other standard logic families (TTL, CMOS, etc).
    In any case, probably wait for a 1-bit paper.

  12. 2 months ago
    Anonymous

    Very hard.
    Prototype probably possible but you would need people for that and it would be hard to prove it's usefulness in order to get actually money from some investor.

  13. 2 months ago
    Anonymous

    Can't you go back into /lmg/ instead of spilling out the papers everywhere?

  14. 2 months ago
    Anonymous

    Here is trinary logic gates in python you can run it in some online compiler

    class TernaryLogicGate:
    def __init__(self, inputs):
    self.inputs = inputs

    def get_output(self):
    pass

    class TernaryANDGate(TernaryLogicGate):
    def get_output(self):
    if -1 in self.inputs:
    return -1
    elif 0 in self.inputs:
    return 0
    else:
    return 1

    class TernaryORGate(TernaryLogicGate):
    def get_output(self):
    if 1 in self.inputs:
    return 1
    elif 0 in self.inputs:
    return 0
    else:
    return -1

    class TernaryNOTGate(TernaryLogicGate):
    def get_output(self):
    if 1 in self.inputs:
    return -1
    elif 0 in self.inputs:
    return 0
    else:
    return 1

    class TernaryXORGate(TernaryLogicGate):
    def get_output(self):
    ones_count = sum(1 for i in self.inputs if i == 1)
    if ones_count % 2 == 1:
    return 1
    elif ones_count == 0:
    return 0
    else:
    return -1

    class TernaryNORGate(TernaryLogicGate):
    def get_output(self):
    if 1 in self.inputs:
    return 0
    elif 0 not in self.inputs:
    return -1
    else:
    return 1

    inputs = [1, -1, 0]
    and_gate = TernaryANDGate(inputs)
    or_gate = TernaryORGate(inputs)
    not_gate = TernaryNOTGate(inputs)
    xor_gate = TernaryXORGate(inputs)
    nor_gate = TernaryNORGate(inputs)

    print("AND gate output:", and_gate.get_output())
    print("OR gate output:", or_gate.get_output())
    print("NOT gate output:", not_gate.get_output())
    print("XOR gate output:", xor_gate.get_output())
    print("NOR gate output:", nor_gate.get_output())

  15. 2 months ago
    Anonymous

    How do make conditionals even in 3 value logic?

    • 2 months ago
      Anonymous

      If x = -1, do one thing, if x = 0 do another thing, if x = 1 do a third thing.

      • 2 months ago
        Anonymous

        I see. Outside of sorting functions I have no clue how to use that. Do you have examples of how common language constructs would be compiled to make use of that?

  16. 2 months ago
    Anonymous

    It's trivial to use ternary states on a binary computer and still both waste very little space AND max out memory bandwidth, which is always the bottleneck anyway. I think the authors meant more native tensor unit support for ternary than returning to actual ternary computer experiments, to avoid wasting energy on arithmetic, and to still maximize BW with funky memory systems.

  17. 2 months ago
    Anonymous

    Soviets built ternary computers in the 50s.
    https://mason.gmu.edu/~drine/History-of-Ternary-Computers.htm

  18. 2 months ago
    Anonymous

    breasts

  19. 2 months ago
    Anonymous

    >How difficult would it be to build hardware that computes in breasts {-1, 0, 1} instead of bits {0, 1}?

    Just subscribe to the quantum computer hype train

    When you're about the nuke a country, you do not ask:

    `Do I do it`,
    `Do I not do it`,
    `Have I died accidentally yet right at the moment of asking this`

    You ask instead

    `Do I do it`
    or
    `Do I not do it`

    Now obviously there is a possibility that ... perhaps you'll die at the moment of the decisions from natural causes, but most likely there will be an operator nearby to continue the execution process

    Therefore asking questions such as

    `Do I do it`
    `Do I not do it`
    `Am I dead`
    Is kind of pointless

    • 2 months ago
      Anonymous

      Moreover, 100% of the times you ask the `Am I dead question` you will be alive, and 100% of the times you fail to ask the question `am I dead`, you'll be dead, so it best to stay at Duality for now, unless you can figure out a better though pattern

    • 2 months ago
      Anonymous

      Moreover, 100% of the times you ask the `Am I dead question` you will be alive, and 100% of the times you fail to ask the question `am I dead`, you'll be dead, so it best to stay at Duality for now, unless you can figure out a better though pattern

      >am I dead?
      What is this schizobabble and what does it have to do with OP?

      • 2 months ago
        Anonymous

        Op wants to use

        1
        0
        -1

        0 and 1 is answer to a question with a yes or no

        0 and 1 and -1 is answer to a question with yes no or maybe

        you can use the the minus one's to naturally check for lack of initialization or whatnot, but overall ....... I do not see the point

        • 2 months ago
          Anonymous

          If you wanna store a maybe(s) you might as well go on with dynamic system like quantum bit measurement, that's why it exist to begin with

        • 2 months ago
          Anonymous

          >0 and 1 and -1 is answer to a question with yes no or maybe
          It's "the answer" to any number of questions you can come up with, like "how much of this factor contributes to the result", which is very relevant for neural nets as per OP's pic.

          • 2 months ago
            Anonymous

            Could you elaborate ?

            So in a scenario where 1 contributes a positive and -1 contributes a negative, why would you record lack of contribution a.k.a 0, is that not a waste of memory

          • 2 months ago
            Anonymous

            >So in a scenario where 1 contributes a positive and -1 contributes a negative, why would you record lack of contribution a.k.a 0
            Because, you fricking moron, your matrix can't have a literal hole in it.

          • 2 months ago
            Anonymous

            The bits as they are of 1s and 0s, need various backups to protect from memory leaks causes by external sources like magnetic cosmic rays, imagine if you make them 1s 0s and -1s, i think you're overly complicating a lot, the notion of sign does not even exist in them as they are

          • 2 months ago
            Anonymous

            Imagine a system where you have to keep 4 separate backups of a byte made up of 1s 0s and -1s, its overly hierarchical

          • 2 months ago
            Anonymous

            I understand nothing of neural nets, other than they are just accumulating weights often percentages slash percentiles slash ratios slash precisions

            They call the `change` the `hidden layer` but to me it looks like some form of a filtering thingy

          • 2 months ago
            Anonymous

            Look at pic related for an illustration of how forward propagation through a neural network relates to matrix multiplication, which is what the actual operation the hardware supports. Then look at OP's pic and

            >So in a scenario where 1 contributes a positive and -1 contributes a negative, why would you record lack of contribution a.k.a 0
            Because, you fricking moron, your matrix can't have a literal hole in it.

            .

          • 2 months ago
            Anonymous

            Look I am not a nerdy guy, nor do I visit IQfy often, in fact I do not ever read scientific notation, either you explain it to me simple or spare it, I do not wanna be a part of another quantum computer hype train

          • 2 months ago
            Anonymous

            >explain it to me simple
            Calculating the output of a neural network involves a series of matrix multiplications. The values in the matrix represent the weights of the connections in the neural network. Those weights represent how much each input contributes to the output for a given layer. Regular neural networks use floats for these values, i.e. 16 or 32 bits per weight. Now someone apparently figured out that most of the time, it doesn't matter exactly how much each factor affects the result, only if it affects it negatively, positively or not at all. Hence -1, 0, 1. 0 is needed because you're still using matrices. You need a value in the matrix for "no contribution". You can't just have a hole there.

          • 2 months ago
            Anonymous

            Okay I am not sure if I understand it properly

            You're envisioning

            3 black and white valued images

            A) cat
            B) human
            C) dog

            Each one is assigned a matrix layer representing blackness to whiteness ratio

            And you're envisioning a dynamic system, in which all the layers begin at 0, 0, 0, but as you make adjustments

            for example cat positive, human neutral, dog negative, a shape begin to take place

            You're also saying that simply firing a ray through the layers a.k.a multiplying them, does not necessarily yield coherent results,

            So if you multiply a cat by a human by a dog, it will not necessarily result in a meaningful black and white weighted image

            But if you choose how much each layer individually matters and contributes to the entire thingy, and do that many times perhaps with slight variations, it might yield more coherent results

            Now two things might have happened here

            1. I am too moronic to grasp what you tried to convey to me

            2. You're just a prompt coomer

            Overall it makes sense for dynamic systems to begin at neutral state, but again, I have no idea if I understood you properly cause I am an intoxicated mess :/

          • 2 months ago
            Anonymous

            Now that this is out of the way

            https://i.imgur.com/khLiPfD.jpg

            How difficult would it be to build hardware that computes in breasts {-1, 0, 1} instead of bits {0, 1}?

            Why `breasts` is this actual patented name for this ?

          • 2 months ago
            Anonymous

            I mean I was already amazed enough to find out about the Sexadecimal, and the Sextuple, its like relics from eternal incel suffering

          • 2 months ago
            Anonymous

            I am currently installing unity stuff, the logo is fricking anal and a pussy

  20. 2 months ago
    Anonymous

    they're called trits

  21. 2 months ago
    Anonymous

    i compute in breasts.

  22. 2 months ago
    Anonymous

    >How difficult would it be to build hardware that computes in breasts {-1, 0, 1} instead of bits {0, 1}?
    Now you just gave me a new project idea. I'm gonna build this thing in verilog, GPL liscence it, and see if AMD will buy the source code to fab it into a chip IRL.

Your email address will not be published. Required fields are marked *