Quantum Computers Calculate Minimum-Energy State of Three Simple Molecules
From <https://arstechnica.com/science/2017/09/quantum-computers-reach-deeper-find-ground-state-of-simple-hydrides/>: Every time we discuss quantum computers, the headline tends to be that someone, somewhere is going to use the quantum to break your encryption and steal your student loan. If only that were true. But it is probably more realistic to think about quantum computers being used to solve quantum problems. ... However, for lithium hydride and beryllium hydride, the solutions are less accurate. The researchers believe that the falling accuracy is not due to the simplifications but hap[p]ens because the qubits do not stay sufficiently coherent during the calculation. In other words, quantum computers are the modern-day equivalent of analog computers, not digital ones: just like in the early days of mechanical- and electronic-based analog machines, they can give answers more quickly, but with less accuracy than, a digital one.
On Fri, Sep 15, 2017 at 08:27:48AM +1200, Lawrence D'Oliveiro wrote:
In other words, quantum computers are the modern-day equivalent of analog computers, not digital ones: just like in the early days of mechanical- and electronic-based analog machines, they can give answers more quickly, but with less accuracy than, a digital one.
No, I think you have fundamentally misunderstood quantum computing. The qubits are a digital bit with an extra state: a mixture (i.e. quantum entanglement) of the two binary states. On measurment a qubit returns only one of two values, binary true or binary false, just like a digital bit. The probability of it returning true (or false) is dependent on the mixture. That's most certainly not analogue: it's something fundamentally different. Cheers Michael.
On Fri, 15 Sep 2017 19:34:08 +1200, Michael Cree wrote:
The probability of it returning true (or false) is dependent on the mixture. That's most certainly not analogue: it's something fundamentally different.
Call it “stochastic analog” or “Monte Carlo analog”, then. “Monte Carlo” simulations are nothing new in the digital world. See also “simulated annealing”. Put it this way: if quantum computers really were digital, then they could be used to solve number-theoretic problems. But it appears they can’t.
On Fri, Sep 15, 2017 at 08:51:07PM +1200, Lawrence D'Oliveiro wrote:
On Fri, 15 Sep 2017 19:34:08 +1200, Michael Cree wrote:
The probability of it returning true (or false) is dependent on the mixture. That's most certainly not analogue: it's something fundamentally different.
Call it “stochastic analog” or “Monte Carlo analog”, then.
And you would be completely wrong. The only outcomes you can measure from a qubit is one of two states: true or false. That's not the case of analogue even if it is stochastic, monte carlo, or whatever.
Put it this way: if quantum computers really were digital, then they could be used to solve number-theoretic problems. But it appears they can’t.
What makes you think a quantum computer [1] can not? Cheers Michael. [1] Here I mean the theoretical conception of a quantum computer just like the universal Turing machine is to a conventional computer. Realising a practical implementation of a quantum computer is certainly proving challenging with current technology, but that is a different issue than whether it is analogue or not.
On Fri, 15 Sep 2017 22:17:53 +1200, Michael Cree wrote:
Realising a practical implementation of a quantum computer is certainly proving challenging with current technology, but that is a different issue than whether it is analogue or not.
There is a fundamental principle that you cannot obtain “something for nothing”. Getting an exponential increase in computing power for a linear increase in componentry would seem to come under that.
On Fri, 15 Sep 2017 22:17:53 +1200, Michael Cree wrote:
The only outcomes you can measure from a qubit is one of two states: true or false. That's not the case of analogue even if it is stochastic, monte carlo, or whatever.
The essence of analog computing is that physical quantities are represented by physical quantities. In previous machines these quantities were considered to be continuous; in this particular example, they happen to be quantized.
On Fri, Sep 15, 2017 at 10:34:14PM +1200, Lawrence D'Oliveiro wrote:
On Fri, 15 Sep 2017 22:17:53 +1200, Michael Cree wrote:
The only outcomes you can measure from a qubit is one of two states: true or false. That's not the case of analogue even if it is stochastic, monte carlo, or whatever.
The essence of analog computing is that physical quantities are represented by physical quantities.
Huh? I'm not sure what you are trying to say here as what you state is just a tautology.
In previous machines these quantities were considered to be continuous; in this particular example, they happen to be quantized.
What previous machines are you speaking of? Cheers Michael.
On Sat, 16 Sep 2017 12:04:27 +1200, Michael Cree wrote:
On Fri, Sep 15, 2017 at 10:34:14PM +1200, Lawrence D'Oliveiro wrote:
The essence of analog computing is that physical quantities are represented by physical quantities.
Huh? I'm not sure what you are trying to say here as what you state is just a tautology.
It’s the difference between analog and digital computers: in an analog computer, a scalar physical quantity in the model being solved is represented by a scalar physical quantity in the machine. In a digital computer, each scalar physical quantity in the machine is just a digit of a number (hence “digital”); you usually have several of these to represent a single quantity in the model, using a place-system encoding, also exponents etc. Thus, physical limitations in the construction of digital machines don’t constrain the accuracy with which we can do calculations, only their speed; whereas in analog machines their accuracy is constrained as well.
On Sat, Sep 16, 2017 at 12:44:20PM +1200, Lawrence D'Oliveiro wrote:
On Sat, 16 Sep 2017 12:04:27 +1200, Michael Cree wrote:
On Fri, Sep 15, 2017 at 10:34:14PM +1200, Lawrence D'Oliveiro wrote:
The essence of analog computing is that physical quantities are represented by physical quantities.
Huh? I'm not sure what you are trying to say here as what you state is just a tautology.
It’s the difference between analog and digital computers: in an analog computer, a scalar physical quantity in the model being solved is represented by a scalar physical quantity in the machine.
Oh, that's what you meant.
In a digital computer, each scalar physical quantity in the machine is just a digit of a number (hence “digital”);
And is only true of the abstract idealisation (the digital model). But in a real implementation (the machine) each physical quantity is represented by an electric potential, a scalar physical quantity. (I hear you say that there is only two values of the electric potential used but that is most certainly not true of real computers.)
Thus, physical limitations in the construction of digital machines don’t constrain the accuracy with which we can do calculations, only their speed;
Only in some limited regime; ultimately the fact the underlying architecture of a modern digital computer is analogue takes over and limits both accuracy (as logic highs are confused with logic lows) and speed (as it takes time for the electric potential to continuously propagate from a logic low to a logic high and vice versa). Pack more transistors into the same space on an IC and the heat generated will lead to thermal fluctuations which leads to a loss of accuracy in calculation even at the same speed. Also the EM coupling between signals increases again leading to a degradation of signal integrity, which appears as a loss of accuracy (i.e. incorrect calculation or processing). These are major problems in implementing dense high-speed digital systems. I think that when you characterise traditional computer architecture as "digital" you judge it by the model and ignore the failings of the machine, and when you characterise quantum computers as "analogue" you judge it by the failings of the machine but ignore the model. Cheers Michael.
participants (2)
-
Lawrence D'Oliveiro -
Michael Cree