nesdev.com

I'm sorry, but I've been hearing about some "top-secret", area 51 type stuff called "quantum computing" and how it's (according to some dumb YouTube video) "100 million times faster" than traditional computing, and they always say that it's because of some magical thing called a "qubit". Now, I'm no physicist or anything, as people act like you need to be one to understand it,

(What the heck is this, a math problem?)



But the somewhat limited definition that I've heard is that "unlike traditional computing where everything is either true or false, in quantum computing, you can have either true or false and both at the same time". Now, unless the people behind quantum computing are God, that's impossible given how they're trying to explain it. I can have a white room, I can have a black room, I can have a zebra striped room, and I can have a gray room, but I can't have a room that's fully white and black at the same time. What I think they're just trying to say is that there's just a third state, like instead of "0" and "1", I have, "0", "1", and "2". Now, if this is true, it doesn't seem very helpful... Is it that as a side effect of being "100 million times faster", it's like this, or does this actually somehow make it "100 million times faster"? Does anyone even really know, because anyone I've heard try to explain it doesn't seem to, and I certainly don't.

It's more or less a parallel computing paradigm, but instead of having two processors side by side you can have two processors "entangled" into one.

Basically it's something that if they can make it cost effective, they can use it to get more power into a smaller space. That's really all it is. Right now it's a long, long way from being cost effective.

So basically, it's a fairly simple concept that everyone is freaking out about as if it's some sort of space alien technology. Nothing out of the ordinary...

The dream is that because qubits combine (entangle) exponentially (i.e. N qubits is equivalent to 2^N bits), there's a lot of potential there, but how to utilize this effectively is an ongoing research question.

To get a more intuitive understanding of quantum computing, you would want to learn about the principles of quantum mechanics: uncertainty principle, probability functions, duality, etc.

It's kind of a messy thing to wrap your mind around but no there is not a third state. Traditional computers can be though of deterministic. A given set of inputs gives the same output given the function evaluting it. Example 1 + 1 always =2.

The results from quantum computers least based on what I know now will always be probablistic. You might get something like 1+1 has a 99.9999% chance of being one.

One big benefit of quantum computing is that it can check more possible input states simultanously making it good for 2^n search type of problems.

I assume you mean "thought"?


I assume you mean two?

Now, I don't have a clue about quantum mechanics or whatever and I don't really care enough about this to learn it, but in what specific scenario would a quantum computer have a more efficient way to find the solution than a regular computer, and how would the quantum computer go about finding the solution vs. a regular computer?

With how rainwarrior said that "N qubits is equivalent to 2^N bits", it seems like this would be useful in itself, because there'd be less data sent for the same amount of information (somehow), assuming it'd be just as easy to move 1 bit vs 1 qubit.

You know, do these special quantum processor things even use transistors? (Actually, are there even logic gates?) Again, if I need to learn about quantum mechanics to even grasp basic principles of quantum computing and how good it is, than just forget it. It's just hard to hear "100 million times faster" and not have questions.

There's a few different schools of thought on it.

One is to create algorithms that are well suited to quantum computation. Like I think the general idea is that reading all the output states is difficult, but if you have a lot of intput bits and only need to read some of the output. There's an quantum integer factorization algorithm, which supposedly would invalidate RSA encryption if it could be run on a suitable quantum computer (no such computer exists yet).

The other school of thought is that eventually you could write a compiler to automatically organize your code into patterns that will fit the quantum paradigm. This would make it a lot more general purpose, and make it a lot more like regular parallel computing. This is an ongoing area of research.

The current state of state of quantum computing is mostly theoretical. We can, of course, already simulate the finished quantum computer with a regular computer (just slowly, i.e. emulation). There are a few real quantum computer implementations in research labs, but they're basically only up to doing computations like 2+2=4 very slowly right now. Still in the "because we can" stage.

You can kinda just take it for granted that computers are going to be faster in the future, quantum or otherwise. It's kind of pointless to give any figures like that. I think current real quantum computers are probably easily 100 million times slower than current low end computer hardware .

The probabalistic thing is resolved by redoing it enough times until you meet your confidence interval. Regular computation also has a probabilistic behaviour, it's just got very high level of confidence. Integer factorization is a well suited task because it's slow to compute but very quick to verify. You can just keep redoing it (i.e. increasing confidence) until verification passes. The idea that the amount of redoing is linear (or someting sub-exponential anyway), but the computational gain with each qubit is exponential, so eventually you overtake it with enough qubits.

The real problem isn't the quantum probability, that's well enough behaved by itself that they can deal with it. The problem is noise and stability, environmental interference, etc. the devices involved are so sensitive that it's really hard to keep the noise from overpowering the signal. They're trying to do computations using individual electrons; pretty tricky stuff.

So, instead of actually solving problems, it just kind of plugs information in randomly, and sees if the random thing it plugged in made the problem true? It's like with something like 602 / 7 = x, instead of taking the time to find out to divide 602 by 7, you'd just plug in random numbers for x? Is there some sort of method to the madness? I mean, if it could somehow tell if what it plugged in was further away or closer to solving the equation than what it did earlier and would compensate? It seems like some sort of weird analogue technology, especially when you started mentioning "noise", like a coaxial cable.


It kind of seems like quantum computing would be better for some things, while regular computers would be better for others. Would a theoretical device from the future likely have both quantum and traditional processors in it for dedicated operations? I mean, something like addition is very simple on traditional computers, unless quantum computers can "guess" so fast that they'd overtake it either way.

It's not exactly what you think. It's more like when they do an operation, say the result has two states. One of those states is that it will read 1 75% of the time, and the other state is that it will read 1 20% of the time. The probability is very predictable from quantum theory, you just have to do the operation enough times to figure out if you're in the 75% state or the 20% state. Figuring out which of these two states it was is equivalent to reading a 1 or a 0 from a binary bit.

The actual quantum calculation has a definite result. The probability stuff is just about how you read that result back. Again, the less you have to read back, the more suitable the algorithm is for quantum computing.


The first practical quantum computing applications, if they ever happen, are going to be very specialized, yes. (That was the first approach I mentioned.)


This is the second approach... With enough qubits and a suitable compiler, theoretically it would overpower the traditional approach. We have to prove first that a physical quantum computer is scalable in a practical way, though.


Anyhow, I don't know why anyone would be excited about quantum computing now, unless they're a researcher. The theory is sound, I think, but the question of whether it can be made practical is very much open. I put it in a similar category with cold fusion.

I have not that much knowledge of quantum computers or any of this, it confuses me too. But I believe it's not a computer in the normal sense of programming and running code, but instead like the Qubit=n^2 bits, that it it's self is just a new era of problem solving machines. Like an FPU in a computer, but it computes on a different level of complexity, and is just a lot faster than normal brute-force which out current computers do today. Instead of programming it to do something, you inject the qubits to be something certain, give it other tangled qubits, and from that you can eventually derive the information needed from the qubit stats to find your answer to the problem in the way a brute-force modern computer could, but not nearly as fast. That's why it gets talked about with encryption, encryption without the keys needs the brute-force stateful information to be de-encrypted which is hard to find, but that's where quantum state computing can make it easy to do calculations. That is an area of computing we haven't done anything besides brute-force on algorithms, so it has lots of potential to further that field in general.

I simply though that it's because "that area of computing" could only be made possible using "brute-force". In fact, I thought just about everything with any form of intelligence operated almost exactly like a "traditional" computer, like the human mind, just very complex. I guess that explains how easy it is to decipher objects based on images and stuff like that and how difficult it is to do math problems that cheap $5 calculators can do in a split second.


I've always heard the term "Floating Point Unit", but I've never had a clue as to what that meant...


So by some freak chance, the computer could actually get the wrong answer for a problem? This seems like it would be extremely problematic in any scenario that isn't a microwave, like if it were used in servers holding credit card balances. It seems so tricky, because you want to do the operation as few times as possible for efficiency, but you also want the data to be correct... From what you're saying though, it does seem like it solves the problem "normally" though, just a bunch of times.


What? Inject the qubits to be something certain? You mean like setting them to create a certain value? Tangled qubits? Are qubit stats just like the state in which the qubit is in, like how a bit can either be 0 or 1?

In fact, in the way you say a bit is either 0 or 1, how would you describe a qubit, if it is even possible in the same way? What would 7 be in "qubits"? Sorry if I'm still not really understanding this...

The possibility of error is inherent in every computing application. There is nothing special about quantum computing in this regard.

There is noise and interference and chosen levels of certainty in all computer engineering. A combination of materials and process is chosen to be as accurate as your application needs.

Internet traffic, hard drives, CDs, transmissions from mars, these things all have error correction schemes, because we expect some amount of corruption.

In a CPU, usually it's not done with an error correcting algorithm, I don't think (though there is such a thing as error correcting RAM), but error tolerance level is still a big factor in design. How close can components be put together, how fast can they switch, etc. all of this affects the reliability of its operation, and there is an engineering decision to allow some % of error to happen. (Yes, the computer you're using DOES make mistakes some amount of the time.)

Quantum computing is exactly the same in this regard, do the operation until you meet your target % of confidence.

On the NES, if we're using DPCM while reading the controller, we simply re-read the controller until we're confident we have valid input.

The potential for error is not really that unusual a problem with quantum computing; at the high level, it's the same problem as other kinds of computers; you engineer the tolerance for what you need, create redundancies, checksums, etc. whatever is appropriate for your application.


The thing that is very unusual about quantum computing is just this potential for exponential growth of power, provided we can ever find a practical way to implement it. (I'm not terribly confident we'll ever get there, but I'm glad that people are looking into it, just in case.)

On my 2 quotes:

An FPU is part of a computer that handles floating point operations. These are just mathematical operations that have a large number of numbers, and sometimes decimal points. They're doable in software, but a hardware FPU to do the mathematics needed speeds them up. Just like Quantum computing can speed up other applications. Just google FPU and start reading, I won't spoon feed it, google is just one website away.

And yes, with today's easy-to-do programming and inexpensive hardware, that area of computing can exist. If quantum computing gets good, it will also eliminate that field. Or at least change it a lot to how it is today.

But yes, you don't just read an atom and get information you need. You have to inject data into them, just like in binary. It's not exactly like that, but the idea is you put data in, wait, data out. Exponentially faster than modern computers is all.