Google claims breakthrough in quantum computer error correction

https://www.ft.com/content/9e04ff7d-f56a-4d5f-a4fa-849336abf659

by aarghh • 3 years ago

128 44 3 years ago

ft.com

ur-whale• 3 years ago

https://archive.is/nwLix

mikece• 3 years ago

Claim != Proof

texaslonghorn53• 3 years ago

You're quibbling over semantics. here's the paper: https://www.nature.com/articles/s41586-022-05434-1

hiddencost2• 3 years ago

Publishing in the most prestigious journal on the planet implies it received some of the highest scrutiny peer review around.

Strilanc4• 3 years ago

This result is notably more open than most papers. The circuits executed and measurement data collected are available on Zenodo: https://zenodo.org/record/6804040 . You can do your own analysis of the claim.

meltyness• 3 years ago

Science! https://www.nature.com/articles/s41586-022-05434-1

texaslonghorn53• 3 years ago

https://www.nature.com/articles/s41586-022-05434-1

https://www.nature.com/articles/d41586-022-04532-4

The FT article is a bit fluffy. here are links to the paper and briefing in Nature.

cubefox• 3 years ago

Relevant quote from the briefing, which indicates they have not yet reached error correction takeoff velocity:

It is known in the field that, when the physical error rate of qubits is high, the probability of logical error increases with increasing system size, whereas when physical error rates are low, increasing the system size leads to the desired exponential suppression of logical error. We feel that we are currently in a ‘crossover’ regime between these scenarios, in which increasing system size initially suppresses the logical error per cycle, but would, with increasing size, later increase error rates. Therefore, it is imperative that we continue to improve both qubit performance and system scale.

So this result is not yet the major breakthrough that would be required to build a scalable quantum computer.

alphabetting• 3 years ago

Official video on the announcement https://youtu.be/_ugJLuJ1_gM

DebtDeflation• 3 years ago

Can it factor 21 without resorting to tricks like precompilation?

Strilanc4• 3 years ago

No, that's years in the future at least. Factoring 21 without any compilation tricks requires doing a modular exponentiation under superposition. The best known way to do that requires two registers of workspace (10 qubits), plus a teensy bit of breathing room (2 qubits), so call it a dozen logical qubits. If all compilation tricks are banned, even the ones that are reasonable for huge numbers but work a bit too well for small numbers such as using small lookup tables to fuse some of the multiplications together, the overall computation takes on the order of 10000 gates. If you require it to work in one or two shots (otherwise even random coin flipping will work), then those gates need to have error rates below one in a hundred thousand and your storage needs error rates per round below one in ten million.

The experiment being announced here is testing different ways storing 1 error corrected qubit, to show that making it bigger can make it better. On an absolute scale, that logical qubit is still not good enough. It needs to be made even bigger. And there needs to be a dozen of them instead of one. And it's barely breaking even; you want strong gains in quality from adding quantity not just minor gains. This means the underlying physical qubits still need more improvement. There's a lot to do!

Disclaimer: am on google quantum team, opinions are my own.

sebzim45002• 3 years ago

Sounds like you are just talking about Shor's algorithm. Presumably if you wanted to factorise 21 naively you could do it with fewer qbits, you just wouldn't demonstrate any kind of quantum speedup.

sebzim45002• 3 years ago

Like all claimed QC breakthroughs, best to wait for Scott Aaronson's response on his blog.

hiddencost2• 3 years ago

Publishing in Nature is a pretty strong signal the results are real. It's entirely possible Scott was one of the anonymous submission reviewers.

I look forward to hearing Scott contextualize it.

dekhn4• 3 years ago

as we used to say in grad school: publishing in nature is a pretty strong signal your results are wrong, but sexy.

texaslonghorn53• 3 years ago

It's entirely possible Scott was one of the anonymous submission reviewers.

Considering that Scott works in theoretical quantum complexity theory, I highly doubt that he reviewed this experimental quantum error correction paper.

throw_pm235• 3 years ago

Or even better: Gil Kalai.

throw_pm235• 3 years ago

In 15 years as a TCS researcher I have followed one heuristic: if something has the word "quantum" in it, I ignore (articles, surveys, conference talks, projects, etc.).

This is more from personal ignorance/laziness and convenience than strong conviction: you cannot follow all areas, you have to make some choices how you spend your time, and this is one particular area that is easy to delimit. (EDIT: and if it does turn out to be a dead-end, I can be glad I made the right call.)

At times this policy has been quite hard to follow, and I may reconsider it sometime, but so far it has served me well.

travisporter• 3 years ago

TCS?

throw_pm235• 3 years ago

theoretical computer science

bordercases• 3 years ago

Theoretical Computer Science

wilsynet• 3 years ago

It would be one thing to say yeah, you don’t think the field is interesting and is likely to be a dead end.

But that’s not what you’re saying. Instead you are saying that you don’t follow it because well, you can’t follow everything. And I agree with that. But in that case, you could go to every single HN topic and post “I don’t follow <insert topic here>, I’m just posting to tell you that I can’t follow everything. So far it has served me well to not follow everything.”

Which doesn’t seem particularly useful or contributory in any way.

birdyrooster• 3 years ago

I just wanted to post a quick comment apologizing for extending this thread any further.

throw_pm235• 3 years ago

Sorry, you are right. I didn't want to post something too negative about it. Obviously, I have to decide what I don't follow, and underlying such a decision is also a belief that the field is more likely to be a dead end than others. But I have no solid argument to back that up.

dekhn4• 3 years ago

Within the field of CS, this is not a bad approach.

However, if you work in: biology, physics, or chemistry: quantum is a frequently used word. It covers far more than QC, entanglement, coherence, tunnelling, or any other crazy bits of quantum. It forms the basis for our atomic theory of matter and has led to extraordinary engineering and science projects.

I used to be excited by DNA computing but it became clear quickly that regardless of any stated advantages of DNA computing, they were tiny compared to the modern working digital computer and the global infrastructure dedicated to improving it year after year (even after Moore's law putters out).

osigurdson• 3 years ago

Interesting that the top rated comment on a post is related to how not reading the post (or any posts in the category) is optimal from a time management perspective.

Ironic that writing this comment as well as reading it is considered a worthwhile time expenditure.

throw_pm235• 3 years ago

But this is quite common -- on most bitcoin-related topics the top comment usually points out (correctly in my opinion) how it is all bogus. I don't have the knowledge to say the same about quantum computing, but my heuristic points in that direction.

meghan_rain• 3 years ago

In 15 years as a TCS researcher I have followed one heuristic: if something has the word "quantum" in it, I ignore (articles, surveys, conference talks, projects, etc.).

Same for "crypto" here :-)

wwarner• 3 years ago

Maybe I missed it, but the ft article claimed the improved error rate was due to improved cooling and better components rather that through better error correction. That’s not what the title says…

abdullahkhalids• 3 years ago

Let me break this down.

* In an error correction code, you encode a logical bit/qubit into a set of physical bits/qubits.

* Error correcting codes come in families, parameterized by integer distance d. Incrementing d, leads to a code with more physical bits/qubits, n, but also the ability to correct errors on a larger number of bits/qubits, j.

* If the error probability on each qubit is p, then on a code of size n, there will be on average n*p errors. It should be immediately clear that if p is small, then n*p<j and the code can correct errors that occur, but if p is large then n*p>j and there will be errors that the code can't correct.

* If the code corrects any physical errors that do occur, then there won't be a logical error (value of logical bit/qubit unchanged), otherwise there will be a logical error. In summary, given a p, you have to pick the right sized code from your family so that n*p<j, and you don't incur any logical errors.

* Another way of saying the same thing is that if p in your hardware becomes small enough, then as you increase your distance d, your logical error rate will go down.

These guys are claiming that their p is small enough that the distance 5 code has a smaller logical error rate then the distance 3 code, which is indeed a breakthrough (if correct). No one has done something like this before to my knowledge.

# Criticism

* The results are limited to storage errors. All they are doing is initializing the logical qubit in some initial state and repeatedly doing error-correction on it, to simulate a qubit at rest while the computation is happening elsewhere on some hypothetical other logical qubits. They have not attempted to do any experiments with applying gates to the qubits. Those will likely yield a much larger error rate. In particular, they are only testing a single logical qubit here, but the interesting gates would be two-qubit gates between two logical qubits, which are necessary to do any non-trivial computation.

* The experiment is limited to 25 cycles of error-detection. This means that their experiment shows that their device could hypothetically implement a depth ~25 circuit. As you might realize, useful circuits have depth many orders of magnitude larger, so this continues to be toy device.

The above is what immediately springs to mind, but I am sure the actual experts will soon chime in. My subjective opinion is that the technical achievements of just running the experiment are very impressive. This is a long journey to useful QCs, but this is nice milestone along the way.

Strilanc4• 3 years ago

In the surface code the gates are all variations on idling. First you idle one way, then you idle another way for a bit, and the result is a gate. (The technical term is lattice surgery.) Because of that, it would be extremely surprising if the gates had notably different error rates from storage. Idling is already a very busy state of affairs.

f0e4c2f7• 3 years ago

A topic I've been interested in lately is quantum machine learning[0]. Qubits as neurons make sense as a natural architecture to me. Though, as I understand we are still somewhat early in terms of the number of available qubits being useful.

While reading about the actual advantages of quantum machine learning over classical machine learning something that came up was a type of error correction you can do in quantum computing that would make backpropagation faster.

Does anyone who understands this better know if this breakthrough might theoretically apply for that application (in the future, with more qubits of course)?

[0] https://en.m.wikipedia.org/wiki/Quantum_machine_learning

Google claims breakthrough in quantum computer error correction

Related Stories