I'm amused by the burying of the non-lede until half way through the paper. I, too, can maintain a 59-bit repetition code for over two hours on my quantum laptop (it's just a normal laptop, but it definitely obeys the laws of quantum mechanics):
Initialize the coded bit, using a 59-qubit repetition code that corrects bit flips but not phase errors, in IPython:
In [1]: A = 59 * (False,)
Write a decoder:
In [2]: def decode(physbits):
...: return sum(physbits) > len(physbits)/2.0
Wait two hours [0]. I'll be lazy and only decode at the end of the two hours, but if I wanted error correction to get the full advantage, I would periodically run the error correction algorithm and fix detected errors. Now decode it:
In [3]: decode(A)
Out[3]: False
Holy cow, it worked!
I'm being rather tongue-in-cheek here, of course. But it's genuinely impressive that my laptop can stick 59 bits into DRAM cells containing a handful of electrons each, and all of them are just fine after several hours. And it's really really impressive that this research group got their superconducting qubits to store classical states well enough that their rather fancy error correcting device could keep up and preserve the logical state for two hours. [1]
But this isn't quantum error correction going FOOM, per se. It's classical. A bit-flip-corrected but not phase-flip-corrected qubit is precisely a classical bit, no more, no less.
The authors did also demonstrate that they could do the same trick correcting phase flips and not bit flips, but that's a tiny bit like turning the experiment on its side and getting the same result. Combining both demonstrations is impressive, though -- regardless of whether you look at the DRAM cells in my laptop as though the level is the Z basis or the X basis, they only work in one single basis. You cannot swap the role of level and phase in DRAM and get it to still work. But the researchers did pull that off on their two-hour-half-life device, and I find that quite impressive, and the fact that it worked strongly suggests that their device is genuinely 59 qubits, whereas no one could credibly argue that my laptop contains giga-qubits of DRAM. Fundamentally, you can do classical repetition using a repetition code, but you cannot do quantum computation with it. You need fancier, and more sensitive-to-errors, codes for this, and that's what the second half of the article is about.
[0] I didn't actually wait two hours. But I could have waited a week and gotten the same result.
[1] The researchers' qubits are nowhere near as good as my DRAM. They had to run their error correction a billion times or so during the course of their two hours. (My DRAM refreshes maybe ten thousand times over the course of two hours, and one can look at DRAM refreshes as correcting something a bit like a repetition code.)
This is perhaps not clear enough, but the title refers to a pattern. For classical bits on a quantum computer this pattern is already playing out (as shown in the cited experiments), and for quantum bits I think it's about to play out.
Classical storage of classical bits is still far more reliable, of course. Hell, a rock chucked into one bucket or another is still more reliable. We'll never beat the classical computer at storing classical bits... but the rock in a bucket has some harsh competition coming.
I should maybe also mention that arbitrarily good qubits are a step on the road, not the end. I've seen a few twitter takes making that incorrect extrapolation. We'll still need hundreds of these logical qubits. It's conceivable that quantity also jumps suddenly... but that'd require even more complex block codes to start working (not just surface codes). I'm way less sure if that will happen in the next five years.
I don’t really expect fancier codes to cause a huge jump in the number of logical qubits. At the end of the day, there’s some code rate (logical bits / physical bits) that makes a quantum computer work. The “FOOM” is the transition from that code rate changing from zero (lifetime of a logical bit is short) to something that is distinctly different from zero (the state lasts long enough to be useful when some credible code). Say the code rate is 0.001 when this happens. (I haven’t been in the field for a little while, but I’d expect higher because those huuuuge codes have huuuuge syndromes, which isn’t so fun. But if true topological QC ever works, it will be a different story.) The code rate is unlikely to ever be higher than 1/7 or so, and it will definitely not exceed 1. So there’s at most a factor of 1000, and probably less, to be gained by improving the code rate. This isn’t an exponential or super-exponential FOOM.
A factor of 1000 may well be the difference between destroying Shor’s-algorithm-prone cryptography and destroying it later, though.
Initialize the coded bit, using a 59-qubit repetition code that corrects bit flips but not phase errors, in IPython:
Write a decoder: Wait two hours [0]. I'll be lazy and only decode at the end of the two hours, but if I wanted error correction to get the full advantage, I would periodically run the error correction algorithm and fix detected errors. Now decode it: Holy cow, it worked!I'm being rather tongue-in-cheek here, of course. But it's genuinely impressive that my laptop can stick 59 bits into DRAM cells containing a handful of electrons each, and all of them are just fine after several hours. And it's really really impressive that this research group got their superconducting qubits to store classical states well enough that their rather fancy error correcting device could keep up and preserve the logical state for two hours. [1]
But this isn't quantum error correction going FOOM, per se. It's classical. A bit-flip-corrected but not phase-flip-corrected qubit is precisely a classical bit, no more, no less.
The authors did also demonstrate that they could do the same trick correcting phase flips and not bit flips, but that's a tiny bit like turning the experiment on its side and getting the same result. Combining both demonstrations is impressive, though -- regardless of whether you look at the DRAM cells in my laptop as though the level is the Z basis or the X basis, they only work in one single basis. You cannot swap the role of level and phase in DRAM and get it to still work. But the researchers did pull that off on their two-hour-half-life device, and I find that quite impressive, and the fact that it worked strongly suggests that their device is genuinely 59 qubits, whereas no one could credibly argue that my laptop contains giga-qubits of DRAM. Fundamentally, you can do classical repetition using a repetition code, but you cannot do quantum computation with it. You need fancier, and more sensitive-to-errors, codes for this, and that's what the second half of the article is about.
[0] I didn't actually wait two hours. But I could have waited a week and gotten the same result.
[1] The researchers' qubits are nowhere near as good as my DRAM. They had to run their error correction a billion times or so during the course of their two hours. (My DRAM refreshes maybe ten thousand times over the course of two hours, and one can look at DRAM refreshes as correcting something a bit like a repetition code.)
This is perhaps not clear enough, but the title refers to a pattern. For classical bits on a quantum computer this pattern is already playing out (as shown in the cited experiments), and for quantum bits I think it's about to play out.
Classical storage of classical bits is still far more reliable, of course. Hell, a rock chucked into one bucket or another is still more reliable. We'll never beat the classical computer at storing classical bits... but the rock in a bucket has some harsh competition coming.
I should maybe also mention that arbitrarily good qubits are a step on the road, not the end. I've seen a few twitter takes making that incorrect extrapolation. We'll still need hundreds of these logical qubits. It's conceivable that quantity also jumps suddenly... but that'd require even more complex block codes to start working (not just surface codes). I'm way less sure if that will happen in the next five years.
A factor of 1000 may well be the difference between destroying Shor’s-algorithm-prone cryptography and destroying it later, though.