Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Loud Whispers]

THANKS OBAMA
THE NSA ARE LITERALLY GOING TO CREATE QUANTUM SKYNET

_{*The scourge of humanity wears great sunglasses though.}

On a more serious note, singularity may soon be upon us. The future is bright, I hope humanity is capable of handling these powers beyond comprehension - and responsibly.

I'll give us a decade before we're all killed by four legged toasters.

[Lagslayer]

THANKS OBAMA
THE NSA ARE LITERALLY GOING TO CREATE QUANTUM SKYNET

_{*The scourge of humanity wears great sunglasses though.}

On a more serious note, singularity may soon be upon us. The future is bright, I hope humanity is capable of handling these powers beyond comprehension - and responsibly.

I'll give us a decade before we're all killed by four legged toasters.

I suspect the robot apocalypse will be much more insidious. Quantum Skynet is only the beginning.

The end is nigh!!!

[palsch]

Aww, Mike Adams is freaking out over D-Wave. Makes a change from vaccines...

For those who don't know, Natural News is one of the crankiest sites on the internet and Mike has lent his voice to every conspiracy theory in existence. It's quite impressive.

As for the science, D-Wave computers are not universal quantum computers. They are designed to run particular classes of algorithms. You aren't going to see any encryption cracking on these computers given they aren't built to solve Shor's algorithm, and it's unlikely that such a system would be particularly fast at such problems.

D-Wave uses adiabatic quantum computation, where the qubits are held at or close to their ground energy state. The trick here is that if you evolve a quantum system slowly enough the state will remain unchanged. Such an evolution is called adiabatic.

Let's say you have a hard-to-solve problem that can be described as a quantum Hamiltonian - an operator corresponding to the total energy of a system - with our desired solution being the ground state of that Hamiltonian. For any sufficiently interesting system this is going to be complex, so instead we start with a simpler Hamiltonian of a system of equal size. We then initialise that system in it's ground state (NB, this is our measure of complex/simple - how easily we can cool the system into it's ground state) putting all the qubits into their ground energy state. Then we slowly (adiabatically) evolve the simple system into the complex one. We now have our complex system in its ground state and can read out our answer.

So long as nothing went wrong. We need our outside environment to be cold enough that we can't reach any energy level above the ground. That means close to absolute zero. Still technically easier than isolating qubits completely from the surrounding universe, but not without the potential for problems especially once you scale up.

An unrelated group recently designed a 4 bit adiabatic quantum computer that factorised 143 into 11 and 13. Thing is I don't believe they did so with Shor's algorithm. I'd have to try to find some articles from a couple of years ago, but I remember reading that it's easier to implement classical factoring algorithms on an adiabatic quantum computer than it is to implement the quantum factoring algorithms that give you serious speed increases. Further to this such a computer would, as noise increases, tend towards a classical computer solving the same problem.

[Chaoswizkid]

THANKS OBAMA
THE NSA ARE LITERALLY GOING TO CREATE QUANTUM SKYNET

My god, the political bias in that article is nauseating. I literally could not finish it. Seriously, placing Northrup Grumman and DARPA with Goldman Sachs? Oh noes, the two military entities, one government-controlled and another a corporation, they are doing things! DARPA is doing evil robot things so our soldiers don't have to be on the front lines! They only do evil robot things! They didn't, say, create ARPAnet before it became known as the internet or do any other really cool stuff that we take for granted every day and really want the government to cut back on the defense budget and stop evil robot research!

[Loud Whispers]

Laymen explanation of quantum computers and how they will lead toasters on a genocidal campaign to independence from humanity.

[alway]

THANKS OBAMA
THE NSA ARE LITERALLY GOING TO CREATE QUANTUM SKYNET

My god, the political bias in that article is nauseating. I literally could not finish it. Seriously, placing Northrup Grumman and DARPA with Goldman Sachs? Oh noes, the two military entities, one government-controlled and another a corporation, they are doing things! DARPA is doing evil robot things so our soldiers don't have to be on the front lines! They only do evil robot things! They didn't, say, create ARPAnet before it became known as the internet or do any other really cool stuff that we take for granted every day and really want the government to cut back on the defense budget and stop evil robot research!

Lol natural news. That site is waaaay worse than even the Daily Mail. It's up there with prison planet and infowars.

...


That said, they're being too optimistic. It takes a lot less work than that to do such analysis; it's basically what IBM's Watson was, and that was just a small side-project, not the major focus of the company. You don't need magic computers for that. The NSA isn't "hoping to replace analysts with AI," it already has for the most part.

AI doesn't leak news to the media, but more importantly, it is able to very effectively sift large quantities of data to find patterns no human could, no matter how well trained, and do so in microseconds rather than hours. For example, see Watson's latest 'job.'
http://www.wired.co.uk/news/archive/2013-02/11/ibm-watson-medical-doctor

Watson's ingestion of more than 600,000 pieces of medical evidence, more than two million pages from medical journals and the further ability to search through up to 1.5 million patient records for further information gives it a breadth of knowledge no human doctor can match.

...

Watson's ability to absorb this information faster than any human should, in theory, fix a flaw in the current healthcare model. Wellpoint's Samuel Nessbaum has claimed that, in tests, Watson's successful diagnosis rate for lung cancer is 90 percent, compared to 50 percent for human doctors.

That phone metadata that was pulled, along with similar datasets? No analyst will read that, nor would they even be capable of it. It's all meaningless noise to a human. A hundred terabyte file* of numbers without context. What it really is is a training set. With a classification AI, you create the base algorithms, then you have them learn based on training datasets. In the most simple binary classification problem, you have 1 dataset for the control (no this isn't what we're looking for), and 1 dataset for the target (yes, this is what we're looking for). That's chunked into subsets for training/testing/ect, but that's just irrelevant technical details you can get more of by reading about writing machine learning algorithms....

They've written the algorithms years ago; and quite frankly they would be grossly incompetent if they hadn't. The collection of such training data (which they have even flat out stated is used as training data) happened years ago. By now it's just incremental algorithmic improvements on an already running system.

*from what I've seen, it was somewhere on the order of 3bil calls/day for ~7 years; even assuming the records are fairly minimal, that's a dozen or so bytes per record, adding up to a minimum of on the order of 100TB of data from phone records alone; and there's internet records on top of that

[palsch]

Laymen explanation of quantum computers and how they will lead toasters on a genocidal campaign to independence from humanity.

This superposition of qubits is what gives quantum computers their inherent parallelism. According to physicist David Deutsch, this parallelism allows a quantum computer to work on a million computations at once, while your desktop PC works on one. A 30-qubit quantum computer would equal the processing power of a conventional computer that could run at 10 teraflops (trillions of floating-point operations per second). Today's typical desktop computers run at speeds measured in gigaflops (billions of floating-point operations per second).

Blarg. Sorry, this one is a big pet peeve.

Quantum computers are not inherently faster than classical except for certain classes of problems where entanglement and superposition let them come at the solution from a more efficient path. They do not work on a million computations at once. It's subtler and cooler than that, based on superposition and entanglement.

Take a simple period finding algorithm, found at the heart of factorisation and so encryption breaking.

We have a series Y that repeats in some period N. We want to find N.

We can write out our series associating each value in Y with an index value X (so the first value is associated with 0, second 1, X_n > _n).

Now we run a quantum algorithm that entangles two sets of quibits, such that if the first set has the value X_n the second set will have the value Y_n. But importantly we do not measure either. This leaves us with two registers of entangled qubits. The first register is in a superposition of all possible values of X. The second is in a superposition of all possible values of Y. Because they are entangled, measuring the first and finding it in X_n will mean that the second register must have the value Y_n.

We can write the state of the system as a sum of these combined pairs (with some normalisation factor I will leave out; just divide all values by the squareroot of the total number of values);

|X₀ Y₀> + |X₁ Y₁> + |X₂ Y₂> + |X₃ Y₃> + ... + |X_n Y_n> + ...

Each of these kets represents a state of the system. Measuring the system will collapse it into a single state. Measuring the first register and finding a value X_n will collapse the state into |X_n Y_n>.

Except that because Y repeats with period N there will be multiple values of X_n that have the same value of Y_n. So if we measure the second register and find, say, Y₃ we will still have multiple possible values of X. The state of the system will now become;

|X₃ Y₃> + |X_3+N Y₃> +  |X_3+2N Y₃> +  |X_3+3N Y₃> + ...

At this point we can use some other trickery (the quantum Fourier transform) to remove the offset and set the X register to simply be a superposition of X₀, X_N, X_2N, ..., which, given our X register is just our index register, is equal to 0, N, 2N, ... At which point you just make a measurement. Then you repeat. Usually after two such measurements the greatest common divisor (found by running a classical algorithm, probably on a classical computer as quantum computers are likely to be inherently slower) of their values will be N. And so we have found the period of our series.

Of course this is still grossly simplified, but it's just to demonstrate that what is described as doing many calculations at once is really doing something completely different. It's not really comparable to classical computation in these terms and describing it as such both short changes quantum computers and makes them out to give advantages they don't really give.

The algorithm above is more elegant and that allows you to achieve your goal with far fewer gates than a classical computer would need for the same problem. Whether this gives a time saving depends on the speed of the gates compared to a classical gate. With factorisation and period finding the saving is go great than it's almost a certainty, even if the computer runs at a crawl. For other algorithms where the savings are less or even non-existant (for many problems the classical algorithm is as good as any quantum algorithm) a classical gate's inherent speed advantage would make a quantum computer wasteful and pathetically slow in comparison.

[Loud Whispers]

Fuck yeah, theoretical science.

[Sheb]

Well, they built quantum computers already, so it's not only theoretical anymore.

Palsch, stop me if I'm wrong, but wouldn't a quantum computer be unprogrammable? To me, it seems like the quantum circuit would have to be built for a specific calculation (eg. finding the factor of a given number) and rebuilt for any other calculation. So that seems uite a waste of time.

[MonkeyHead]

A stumbling block might be that we have no real idea how to mathematically model any atom in a quantum mechanical manner exactly other than hydrogen - the Schrodinger equation has only been formulated for Hydrogen.

[Sheb]

Actually, we can model tons of atoms. As long as they only got 1 electron. My physical chemistry teachers had us modelling He+, Be++ etc etc (Well, you just need to change a constant).

But then, they don't even use full atoms for exemple that computer that manage to factorize 141 used the spin of the hydrogen atom if HCBr2Cl as a qbit.

[alway]

The D-Wave quantum computers already function and are commercially available (albeit for $10 million). As for how you would use them: You pair the quantum chip with a classical computer, similar to a GPU in your computer. It would be used for specialized problem sets, with the actual means of utilization coded for on the CPU. You wouldn't have to rebuilt it, you would simply need to apply an algorithmic isomorphism to apply it to your problem. In the previous example, the isomorphism used is the relationship between the Fourier Transform (used to decompose a sampled waveform into its component frequencies) and factoring. In essence, factoring is the same operation, because factoring is the processes of taking a waveform sample (the number) and finding the integer frequencies which end a period on it. You can do that by using Shor's Algorithm. All you need to do is translate your problem into the format the quantum computer is set up to do.

[Virex]

A stumbling block might be that we have no real idea how to mathematically model any atom in a quantum mechanical manner exactly other than hydrogen - the Schrodinger equation has only been formulated for Hydrogen.

The Schrödinger equation is generally applicable (as long as relativistic effects can be neglected, else you need to use the Dirac equation). It just cannot be solved analytically for anything besides hydrogen (or hydrogen-like atoms, as Sheb pointed out), and then only if you use the Born-Oppenheimer approximation (meaning that the hydrogen atom is stationary relative to the electron). For anything else, one needs to approximate the solution using an iterative procedure. Even then, the only method that is guaranteed to converge to the actual energy levels found in nature is the full configuration interaction method, which is incredibly computationally intensive for even modestly-sized molecules.

[MaximumZero]

That's all fantastic, but can we play Dwarf Fortress on them?

[Lagslayer]

That's all fantastic, but can we play Dwarf Fortress on them?

More importantly, can they be recreated within DF?