# robust nine-quantum-bit quantum computer made

humy
Science 09 Mar '15 17:40
1. 09 Mar '15 17:407 edits
This is still far from making a useful practical quantum computer but:

http://physicsworld.com/cws/article/news/2015/mar/09/how-to-make-a-tougher-quantum-computer

it uses "parity" to do quantum error correction allowing 9 quantum bit processing. To make a 'useful' quantum computer, need to do it for a much greater number than just 9 quantum bits while somehow maintaining adequate error correction.
2. DeepThought
09 Mar '15 19:51
Originally posted by humy
This is still far from making a useful practical quantum computer but:

http://physicsworld.com/cws/article/news/2015/mar/09/how-to-make-a-tougher-quantum-computer

it uses "parity" to do quantum error correction allowing 9 quantum bit processing. To make a 'useful' quantum computer, need to do it for a much greater number than just 9 quantum bits while somehow maintaining adequate error correction.
Each qubit is a small circuit consisting of a capacitor and a Josephson junction, and is made from an aluminium film evaporated onto a sapphire substrate.
Sounds expensive! That's really clever. They do the error detection using an observable that commutes with the Hamiltonian, and claim to be able to correct the error. Error correction codes can get quite complex. This is a big step forwards as it means that the machines don't have to be quite as robust. I wonder how many qubits are available for computation.

They say they can cope with errors in two qubits. For an error in one classical bit you can use:

abc def

d = P(a, b);
e = P(a, c)
f = P(b, c)

if a flips then d and e change but f doesn't - so you can tell which bit's flipped. The catch is that if two data bits flip that can't be distinguished from one bit flipping in the error bits. You need as many error correction bits as data bits, and a low probability of one bit flipping, so that the chances of two flipping are very small. They seem to have used a different system, but it leaves me wondering how many data bits they have.

I need to read the full article, which Nature will let me do, one doesn't need a subsciption for this one, but I can't download a copy to my machine.

Bear in mind, Intel's first microprocessor had 4 bits, the 4004. It was a commercial success used in calculators. To do a 16 bit integer add you needed to do four 4 bit adds and cope with the carry. Maybe someone has a use for this thing - well me for one - it's a hobbyists dream.
3. 10 Mar '15 05:29
Originally posted by DeepThought
Bear in mind, Intel's first microprocessor had 4 bits, the 4004.
Err, no, it most definitely did not. It had a 4-bit wide instruction set. Not the same thing at all.

My current PC is running what is called 64-bit Windows. Don't make the mistake of thinking my CPU has only 64 bits.
4. DeepThought
10 Mar '15 06:594 edits
Err, no, it most definitely did not. It had a 4-bit wide instruction set. Not the same thing at all.

My current PC is running what is called 64-bit Windows. Don't make the mistake of thinking my CPU has only 64 bits.
If it had a 4 bit instruction set there'd have been 16 instructions. It had an 8 bit instruction set, with instructions loaded over two cycles, and a set of 4 bit registers. As I understand it, which is minimally, quantum machines don't copy the qubits out of the registers, the operations are done on the bits in the registers. So a quantum machine only needs the one set of registers. On conventional machines the data is copied out of the registers, goes to the execution units, and is then written back. So I don't think you've picked up on any kind of meaningful error in what I said.
5. 10 Mar '15 08:54
Originally posted by DeepThought
So I don't think you've picked up on any kind of meaningful error in what I said.
You said it had 4 bits. It didn't.

As I understand it, which is minimally, quantum machines don't copy the qubits out of the registers, the operations are done on the bits in the registers.
It depends on the design. It also depends on what type of computations you are trying to do with it. Also it works in tandem with traditional computing circuitry so it can't really be compared that easily with early CPUs.

So a quantum machine only needs the one set of registers. On conventional machines the data is copied out of the registers, goes to the execution units, and is then written back.
Nice how you just brush over the execution units. It is the execution units where the action happens, not the registers. The registers are somewhat irrelevant. How many transistors are required to add two numbers?
How many qubits are required to do a basic quantum calculation?
Those are the questions you should be comparing, not how wide the instruction set was.