IBM quantum computers now finish some tasks in hours, not months | Hacker News Reader

IBM quantum computers now finish some tasks in hours, not months

engadget.com

145 pointsnkjoep57 comments

Note that this is not about quantum speedup, but, rather conversely, making the classical part of the computation more efficient. Still important to get these quantum-adjacent things correct, but not as groundbreaking as one could hope for just reading the headline.

I am pretty impressed. Also, this makes digital currencies even more suspect.

AS I understand it, Quantum computers as of yet are only useful for a relatively small set of obscure operations. They're extremely fast of course, but they're not a drop in replacement for basic gate operations that classical computers and their crypto algorithms use. A cryptocoin like bitcoin uses multiple algorithms making them an exceptionally difficult problem to program a quantum computer to handle. If you ask me, we'll hear about Quantum computers breaking a lot of other things before they get to crypto coins.

If quantum computers break regular encryption, then the sudden collapse of Bitcoin would be the literal last of our worries. The inability to buy things using computers securely would be a much bigger problem, especially given how computerized banking is.

Good luck getting cash when your local bank is trying to figure out procedures that haven’t been used in a literal generation.

Bitcoin collapsing would be funny, in hindsight. Probably a pretty emotional event for some in the moment.

But banking is already not that secure. Credit cards suffer from massive ongoing fraud. Especially in places like the USA where chip systems are less common.

There is still a pretty sizable difference between “a patchwork system that fails a lot in small ways” and “you can’t use a computer for anything financial, ever”.

I get the impression that the community would have enough consensus for a snapshot at a certain block number, and continue the chain with new quantum resistant keys

I think for the attacker with a quantum computer, they will need to steal low amounts for a long time so that people brush other user’s missing funds off as user error

But these investigations often do quickly give users the benefit of the doubt so I would see the network being at high alert really quickly

Chip is fairly common nowadays in the USA.

Fraud liability was shifted from card brands to retailers if they don't support chip. That was a big deal to start supporting it.

Would it be any more complicated for banks than enabling this [0] (or whatever comes after it) in their webservers and asking visitors to use Chrome?

[0]: https://security.googleblog.com/2016/07/experimenting-with-p...

Yes, by several orders of magnitude. The problem isn’t just browsers, it’s the massive pile of code that runs all these systems on the backend, many of which run on honest to god mainframes. You’re not going to flip a switch and get decades worth of software to run quantum proof encryption.

Wouldn't it just need to be the network transport that needs to be updated? I don't imagine the mainframe databases are encrypted-at-rest as it is, so how would quantum change anything there?

Yes, it’s mostly network transit that’s the problem. The issue is that there is a lot of network transit behind the scenes that’s not going through nginx and a client browser.

This is specifically what you shouldn't be worried about just based on the speedup that the article focuses on. This is about a constant-factor speedup from a better-engineered classical component, not the exponential speedup you get from being able to handle additional qubits.

Engineers are already working on quantum-proof blockchain technology.

I love it...I think that's got to be peak vaporware

It needs lighting speed factorization to break Bitcoin. The public key of a bitcoin address is exposed only when the transaction is broadcasted on the network. The attacker would have about 10 minutes to factorize the public key before the transaction is validated and the original address's balance is cleared.

Factorization applies to RSA. Bitcoin uses ECDSA, so you'd need to solve the elliptic curve discrete logarithm problem.

You're right.

This is not true: most value is stored on hardware wallets, and the public keys are exposed to the computers whenever the balance is checked. It’s dangerous to treat public keys as private information, as most software is not designed to protect it.

A bitcoin address is the double hash with two different algorithms of the public key. Hardware wallets do not expose the public key to check the balance, they use the address. The public key is broadcasted when a transaction is signed so the network can verify the signature. This is the reason why bitcoin addresses should never be reused. It gives the attacker the information and the time to break the encryption (theoretically).

The address is the double hash of the hierarchically derived public key using BIP32 key derivation. To be able to compute multiple addresses the master public key is used in the computer's memory that may contain malicious code.

Qiskit has been out forever (https://qiskit.org/). What do they mean by "plans to release it in 2021"?

If anyone is interested - my old team published a primer on quantum computing and using Qiskit that should be quite appropriate for a developer audience. Code samples start at page 48.

https://arxiv.org/abs/1907.11513v1

I have no involvement now and have moved on to a different company. The team has also gone on to do way bigger and better things and has published some additional papers that are way over my skill level and knowledge.

Thanks for sharing

v1.0 release?

"Qiskit and improved hardware will lead to a day when anyone can put quantum computing to use, even if it’s through a distant mainframe." J. Fingas, Engaget, 2021

"I think there is a world market for maybe five computers." Thomas Watson, president of IBM, 1943

> "I think there is a world market for maybe five computers." Thomas Watson, president of IBM, 1943

Watson was not wrong, even though everybody likes to quote this and act like he was dumb.

At the time he was saying that a computer was the size of a 2 story building and probably cost half a billion dollars in today's money (maybe even more?) plus who knows how much in maintenance costs.

And their applicability was extremely limited plus potential customers didn't even realize that they'd need them.

What's funny is we might get there someday with enough server-side push and wearable/mobile/PC tech pushing computation to the cloud and streaming it "down" to nimble devices.

EDIT: After further thought, I realize this is the point OP is making.

"I think there is a world market for maybe five cloud computer companies" — what Thomas Watson meant

Even that quote would be ridiculous. There must be 10,000’s serious cloud providers.

So, AWS, Azure, GCP, and two more?

DO, Linode

So, there is a world market for five quantum computers that anyone can use?

That IBM quote from Watson is fictional.

I think this year will IBM factor 55 using Shor's.

Genuine question, is this sarcasm or reality? IBM's marketing of things like this is highly questionably, especially since blockchain and watson turned out to be such smokescreens from the marketplace/product/adoption standpoint.

1000 qubits in two years seems like a "quantum" leap based on what I read.

In 2019 ibm tried 35 but there was too much noise, I think this year they will skip 35 and go directly to 55.

It doesn't seem likely that they'll factor 55 within 3 years

>Genuine question, is this sarcasm or reality?

Pretty sure it's both.

The hype level now being applied to all things "quantum" makes the hype that was applied to AI a few years ago seem like mere puffery.

IMHO thats a bit blinkered, i have a deep interest in the field and the only mainstream publications on it clearly note there's no practical applications yet, and we're looking at a 5-10 year horizon of _maybe_ finding a niche application that actually improves with a handful of qubits

Contrast that with AI, which caused formation of new departments and hiring binges throughout industry, including industry far outside tech

And all those hordes of "AI" "engineers" are less useful than a few hundred statisticians. If the buzzwords get innumerate business types to invest in building models, that is great. I just wish they would let people use things that actually work and that a human can understand, like linear regression.

We have seen legitimate revolutions in natural language processing, computer vision, and games. Those things, however, are still a long way from being useful.

Personally, I've become convinced that AI is just an elaborate jobs program scheme thought up by washed out PhD students to bilk gullable venture capitalists out of endless amounts of money.

> 1000 qubits in two years seems like a "quantum" leap based on what I read.

It is; that's enough to make a single error-corrected qubit. So the leap here is effectively one quantum

Those SOBs were too busy planning the demise of CentOS to bother

Very confused why they're not trying 42 first.

Could be because 42 is not a semiprime.

Even given the caveats about what was actually sped up -- coming soon:

"It's 2022 and I cannot believe we are expected to wait hours for this process to finish."

what are the implications with regard to encryption and crypto, if any?

This is only about speeding up the support stuff. No actual quantum improvement is discussed.

There is a tremendous amount of misinformation floating around on this topic. The short answer is that we are not even close to starting yet. This has a good discussion:

* https://crypto.stackexchange.com/questions/59795/largest-int...

Note the new and exciting estimate of only 20 million physical qubits required to crack RSA 2048. The article mentions the hope of achieving 1000 physical qubits.

https://en.wikipedia.org/wiki/Transistor_count

1,000 on a chip in 1970. 20M by the late 1990s or so.

None at all.

Does "crypto" refer to the branch of mathematics or to the scam?

It has substantial theoretical implications in cryptography, but any practical applications are decades off.

Cryptocurrency couldn't get any more broken, so it has no implications for that.

> Cryptocurrency couldn't get any more broken

So you have a way to steal Bitcoins?

1) Am I hitting a paywall or is the article really only 4 paragraphs long, amounting to a directly-relayed press release?

2) Is it more likely that this is funded by the NSA, or instead that the NSA has had something similar for decades and this is a ploughsare-type filtration of techniques into the private sector?

What does the phrase "ploughsare-type filtration of techniques into the private sector" mean?

I'm thinking it's this: https://en.wikipedia.org/wiki/Swords_to_ploughshares

1) Yes

2) No, ploughsare filtration techniques seem to go in reverse like self driving cars.

Reply on news.ycombinator.com