Google expected to achieve quantum supremacy in 2019 | Hacker News Reader

Google expected to achieve quantum supremacy in 2019

thenextweb.com

58 pointskurthr27 comments

I used to do ion-trap based quantum computing research but I've been out of the field since 2009. It's great there's multiple large entities investing in quantum computing. Honestly I had been getting pretty skeptical on any of the approaches (ion tap, superconducting, neutral atom traps etc...) because it seemed like little progress was being made.

I wish I wish there was more here to go on. There's nothing 'new' here in the actual article I can see, except an extrapolation from a trend observed for VERY low #'s of qubits forward; seems like the shakiest kind of speculation.

Quantum simulations see like the lowest hanging fruit for useful computation out of these low-qubit systems; I look forward to seeing if they can use their system to solve some currently intractable problems!

I took some courses on quantum computing in college, and my takeaway at the time was all modern approaches to quantum computing are slower (effectively, and worse in almost every case in terms of algorithmic complexity) than their classical counterparts. Are we any closer to getting a quantum implementation of Shor's algorithm? My understanding is that none of D-wave's systems for instance are anywhere close to implementing Shor, and I treat that as my "are we there yet" test. I really enjoyed the classes, but haven't kept up with advancements in the space in really any way.

> My understanding is that none of D-wave's systems for instance are anywhere close to implementing Shor, and I treat that as my "are we there yet" test.

Do you care about the algorithm, or the result? Currently we can factor 11-bit semiprimes fairly reliably, and our hardware is capable of running 16-bit problems. But no, it's not Shor's algorithm. It's conceptually much simpler: implement a multiplication circuit, clamp the output, anneal to find the inputs. Despite not being Shor's algorithm, we've yet to see a competitor demonstrate anything close on the factoring problem.

There's a strange thing in quantum computing happening right now. Certain people believe that there's only one kind of quantum computing, because it's got a provable speedup under yet-unobtainable assumptions. Adiabatic quantum computing, under a similarly unobtainable set of assumptions, can run the same algorithms, polynomially equivalent in time and space. But gate-model gets all the hype.

And speaking of hype, 2017 just called. https://spectrum.ieee.org/computing/hardware/google-plans-to...

I mostly care about the available time for Bitcoin to switch to quantum secure cryptography. It’s scary that people wouldn’t accept a 10x fee increase which would be necessary for the switch.

It's already too late. Even if they switched now it wouldn't save all the coins stored in older addresses.

They need to choose between removing a large part of the coins from the blockchain or accepting that those coins will be taken by whoever has the technology first. Both options are bad for bitcoin.

It'a not that bad, it's easy to add a new digital signature to Bitcoin (the hard part is developing one that's accepted by the community). Most of the money can be consolidated to less UTXOs. The problem will still be that we have no idea when the current signatures get unsecure.

Are there any resources you can recommend to understand D-wave's quantum computing a bit better?

I took a very basic course about gate-model quantum computing at my university. The (mathematics) professor would have loved to be able to explain adiabatic quantum computing on a basic level, but was unable to find entry-level material to really understand how it works or what problems it can solve.

It doesn't matter if classical solutions are better at this point. We need to make progress and any baby step helps.

According to some sceptics it does matter. Some are saying that error rate will grow the same way as “computational power” ie. you can’t physically get value out of quantum computing and it will never outperform classical computer. I’m not saying that is the case but the argument is sound and until proven otherwise we won’t know.

Error rate remains more or less proportional with qubit growth as long as companies optimize for qubit growth alone. But once you have thousands of qubits you can use that to build something like a few dozen error-free logical qubits. Most companies don't believe it's worth attempting that yet and that it's better to optimize for qubit growth for now.

I think that's what they are arguing with - if you increase number of entangled qubits - your error rate will explode the same way that your "computational potential". If that's the case then the whole quantum computing has a huge problem, you'd have to cool it down closer and closer to absolute zero. Having more machines that do poor calculation won't help here - you'll end up with a farm of expensive fridges. Just for the record - I want this to be false, quantum computing will hopefully work.

Expecting D-wave to get shor working on its quantum computer is like expecting Nvidia to run Windows on its gpus. D-wave isn't even attempting to solve problems like the shor algorithm.

Reading the current Science article on this topic I got the impression that ‘quantum advantage’ might be a more reasonable first goal. that is we are still waiting for a quantum computer which can perform better than an ordinary computer in a single restricted setting. Also I was a little bit shocked that even with 16 Qbit those systems already show a significant error rate (I think up to 50%)

My skepticism was confirmed by estimates in the same article of the Quantum supremacy still being decades away. So I’m a little bit confused about the current announcement but fortunately it is easily falsifiable... let’s wait for 2020

Let's say I'm not Google or IBM. Once the seal breaks, so speak, what type of resources will I need to get involved in quantum computing? And what are some of the possible nefarious things I might dabble in?

Is quantum computing going to replace nuclear weapons as an uncontainable force to be feared?

> what are some of the possible nefarious things I might dabble in?

My concern is breaking today's crypto so only the big kids can have it. I've no idea if this is a real worry -- lot's of smart people here could probably address that....

Being quantum-safe is mostly a matter of switching to hash functions and doubling the sizes. A total reworking but not a permanent break.

But all your old stuff will be basically public.

As someone I knew pit it: no one cares about financial stuff from two years ago. Or even last year. But all your healthcare and private data is very very interesting!

It's not that interesting though. I mean, it's not 100% worthless information, but it's not as valuable as people pretend. Especially if the legal environment continues in the direction it is going today, private data (especially medical data) becomes pretty devoid of monetary value if not legally obtained, and even becomes a liability. The threat of Big Companies spending billions on cracking your encryption for filthy lucre is pretty much an unrealistic fantasy.

Maybe I'm missing something, but it appears to me quantum supremacy cannot be proven empirically, only theoretically. Quantum supremacy means you can prove that a quantum computer performs task A faster than any classical computer could do it. Not faster than a classical computer currently does it. For example, Schor's algorithm can factorize in polynomial time and classical factorization takes superpolynomial time. The problem is that it's not easy at all to show classical factorization is superpolynomial, and some people actually think there's a high likelihood it's actually polynomial. So, quantum supremacy is pretty much theoretical CS in the area of lower bounds for classical algorithmic problems.

If that's true then why not keep it secret for a few more months until you are confident to release it outright? This is definitively going to be underwhelming.

Supremacy is the word that seems awkward here. "We might have something to show by the end of the year" seems to fall a little short of supremacy.

And there is something about a double exponential rate? It still seems like we have a double exponential rate of "might show something".

Of course I do look forward to seeing the quantum computer playing chess or something.

Quantum supremacy is actually a more specific concept than you might be thinking: https://en.m.wikipedia.org/wiki/Quantum_supremacy

There is such a thing as a double exponential rate:

https://en.m.wikipedia.org/wiki/Double_exponential_function

If their claims are correct, using the Wikipedia example, if they are as fast as a cell phone in year one, then they should be faster than something like a million of the biggest super computer in year two.

In year three, the incremental improvement will be incomprehensibly faster.

So, they are making some pretty strong claims. Their current claims are definitely stronger than the ones that led to the AI winter.

(edit: and, if it could do physical simulation poorly in year one, by year four or five, it would be simulating an incalculable number of universes starting from the Big Bang, at much faster than real-time. I guess they could then recursively contain comparable quantum computers, which would progress even faster than the host machine?)

As shown in the graph in your link, the double exponential has a sharper "elbow" and therefore it seems to me that if you start at the same y-value, then the double exponential will lag behind the single one for a period of time, which in practice might be any length.

Some boob at google's QC group overfit 3 data points, two of which were questionable, and called it a "law" named after himself. I look forward to people attempting to justify why this is still a "law" when there is still no quantum supremacy 10 years from now...

>Nevin’s Law is currently more of an affectionate term for a rule coined by Google‘s Hartmut Nevin. At the company’s Spring Quantum Symposium this May, Nevin made the claim that quantum systems are increasing in performance at a doubly-exponential rate. This means, rather than doubling in performance with successive iterations as was the case with classical computers and Moore’s Law, quantum technology is increasing in performance at a much more dramatic rate. It took 50 years to go from punch card systems to iPhones: if Nevin’s Law is true we’ll see quantum systems increase in a fraction of that time.

What a character!

Somewhere out there... Skynet is gestating. What an exciting time to be alive.

Reply on news.ycombinator.com