Q-Day - let's hope we're preparing for it

[dyqik]

We know what we can do, and we can do it easily. We can increase the key length on TLS RSA encryption used by webservers, and use alternative algorithms that are already built into TLS and other systems like that that are harder for quantum computers to break. Even more algorithms are already built in to libraries available to the software developers. Rolling this out will take far less time than developing quantum computers - probably less time than getting people to shift to https from http.

For high security applications, we can use quantum encryption - this will take longer to develop and roll out, but it's already in use.

Quantum computers aren't a pandemic that will go from zero to millions in a few months.

And "generative AI" hasn't really gone anywhere. What's come out is just autocomplete, with zero understanding and zero intelligence. It's good for generating spam and it's equivalents, and not much else.

[shpalman]

The point of the article is scaremongering clickbait from someone who by his own admission doesn't know what he's talking about and spends nearly half the article waffling about something else entirely in order to meet his word count.

Right now I am actually watching an online talk about quantum computing for electronic engineers.

This just got referenced: https://sam-jaques.appspot.com/quantum_landscape_2023



You can actually look at the 2021 and 2022 updates to see the progress. And then remember that the graph has logarithmic axes.

At https://quantumalgorithmzoo.org/ the first entry talks about the algorithms which would break certain kinds of encryption if you could implement them.

If you can't really see what use the other algorithms would be, well, the algorithms are there because a quantum computer would be able to do them, not because anyone needs them to be done.

[shpalman]

And this is the linear-axis version, "the distance from today's quantum computers to breaking RSA is about 10,000 chart-widths."



The 2022 update says

IBM released data from its 127-qubit Eagle processor. I have a cynical hypothesis: it wasn't quite ready to demonstrate when they announced it last year, but they had to make an announcement since they had just made their quantum roadmap in 2020, and it would look really bad to fall behind only one year after making the plan. I'm still impressed, even if it's slightly late. They have also announced a new 433-qubit chip, which I have again excluded from the chart because there is no data on qubit quality yet.

This post about Eagle from March 2022 says "IBM Quantum is committed to following our roadmap to bring a 1,121 qubit processor online by 2023" but, well, it's briefly mentioned in this press release from December 2023 while they'd prefer to talk about something apparently useful done with 127 noisy qubits in a way which wouldn't be feasible on a normal computer: Evidence for the utility of quantum computing before fault tolerance.

They simulated a physical system consisting of interacting two-level systems in a 2d network... using a quantum computer based on interacting two-level systems in a 2d network.

And here's Sabine explaining a classical simulation that is significantly more accurate and precise than the results obtained from the quantum processor.

[dyqik]

Ah, good old fashioned analog computing

[shpalman]

... a classical simulation that is significantly more accurate and precise than the results obtained from the quantum processor.

... we compute the expected value of the average single-site magnetization and show that we can obtain an accuracy of approximately
10^-14 with a simulation that runs in less than 10 s on a laptop computer...

... we obtain values of these higher weight observables to orders of magnitude better accuracy than the quantum processor with a simulation that takes less than 4 min (for these observables) to run on a laptop and a state that takes up, at most, 0.3 GB of memory.

Groups will continue to claim "quantum advantage" in solving extremely contrived problems and not long afterward someone will come up with a way of doing it on a laptop in minutes.

[jimbob]

The point of the article is scaremongering clickbait from someone who by his own admission doesn't know what he's talking about and spends nearly half the article waffling about something else entirely in order to meet his word count.

Right now I am actually watching an online talk about quantum computing for electronic engineers.

This just got referenced: https://sam-jaques.appspot.com/quantum_landscape_2023



You can actually look at the 2021 and 2022 updates to see the progress. And then remember that the graph has logarithmic axes.

At https://quantumalgorithmzoo.org/ the first entry talks about the algorithms which would break certain kinds of encryption if you could implement them.

If you can't really see what use the other algorithms would be, well, the algorithms are there because a quantum computer would be able to do them, not because anyone needs them to be done.

What does "surface codes work" mean in this context?

[shpalman]

A surface code is an encoding of a qubit states in a two-dimensional grid of physical qubits. Beyond that I don't really know what's going on with surface codes, maybe I have the recording of the IBM lecture which mentions them. Maybe that ESRF lecture I linked to above has something about them.

[shpalman]

... a classical simulation that is significantly more accurate and precise than the results obtained from the quantum processor.

... we compute the expected value of the average single-site magnetization and show that we can obtain an accuracy of approximately
10^-14 with a simulation that runs in less than 10 s on a laptop computer...

... we obtain values of these higher weight observables to orders of magnitude better accuracy than the quantum processor with a simulation that takes less than 4 min (for these observables) to run on a laptop and a state that takes up, at most, 0.3 GB of memory.

Groups will continue to claim "quantum advantage" in solving extremely contrived problems and not long afterward someone will come up with a way of doing it on a laptop in minutes.

Or on a Commodore 64.

https://www.tomshardware.com/tech-indus ... y-accurate

[shpalman]

Replication of Quantum Factorisation Records with an 8-bit Home Computer, an Abacus, and a Dog (pdf)

New technologies, when introduced, are typically given names that overstate their capabilities, usually by equating them with existing familiar systems or technological artefacts. For example the first computers in the 1940s and 1950s, often little more than glorified electric adding machines, were nevertheless described as “electronic brains”. More recently, large language models (LLMs) have been touted as “artificial intelligence”, and complex physics experiments have been touted as “quantum computers”. In order to avoid any confusion with actual computers like the VIC-20 with which they have nothing in common, we refer to them here as “physics experiments”. Similarly, we refer to an abacus as “an abacus” rather than a digital computer, despite the fact that it relies on digital manipulation to effect its computations. Finally, we refer to a dog as “a dog” because even the most strenuous mental gymnastics can’t really make it sound like it’s a computer.

As has been previously pointed out, the 2001 and 2012 quantum factorisation records may be easily matched with a dog trained to bark three times [35]. We verified this by taking a recently-calibrated reference dog, Scribble, depicted in Figure 6, and having him bark three times, thus simultaneously factorising both 15 and 21. This process wasn’t as simple as it first appeared because Scribble is very well behaved and almost never barks. Having him perform the quantum factorisation required having his owner play with him with a ball in order to encourage him to bark. It was a special performance just for this publication, because he understands the importance of evidence-based science.

The process was then repeated to have him bark five times, factorising the number 35 and thereby exceeding the capabilities of the quantum factorisation physics experiments mentioned earlier.

[shpalman]

... The 2022 update says

IBM released data from its 127-qubit Eagle processor. I have a cynical hypothesis: it wasn't quite ready to demonstrate when they announced it last year, but they had to make an announcement since they had just made their quantum roadmap in 2020, and it would look really bad to fall behind only one year after making the plan. I'm still impressed, even if it's slightly late. They have also announced a new 433-qubit chip, which I have again excluded from the chart because there is no data on qubit quality yet.

This post about Eagle from March 2022 says "IBM Quantum is committed to following our roadmap to bring a 1,121 qubit processor online by 2023" but, well, it's briefly mentioned in this press release from December 2023 while they'd prefer to talk about something apparently useful done with 127 noisy qubits in a way which wouldn't be feasible on a normal computer: Evidence for the utility of quantum computing before fault tolerance.

They simulated a physical system consisting of interacting two-level systems in a 2d network... using a quantum computer based on interacting two-level systems in a 2d network.

And here's Sabine explaining a classical simulation that is significantly more accurate and precise than the results obtained from the quantum processor.

In the meantime, IBM have changed their roadmaps. They are no longer talking about increasing the number of qubits on a chip, but rather talking about increasing the number of gates. But gates aren't physical devices (as they are in digital logic) but rather they're the basic qubit operations. So what they're saying is that they hope to increase the number of operations they can carry out on a chip in a single computation.

So when it says "Nighthawk (5K)" they mean (check the small print) 5k gates on 120 qubits.

https://www.ibm.com/quantum/hardware#roadmap

The 2022 version of the roadmap says that in 2023 there'd be Condor with 1121 qubits, and by 2025 there'd be "Kookaburra" with 4158+ qubits. 2026+ said "Scaling to 10K-100K qubits"... The current version doesn't mention Condor's qubit counts but just says "Single-system scaling and fridge capacity" and Kookaburra's qubit count isn't mentioned either, but it's "on target for next year" in terms of "Demonstrate a complete module consisting of a logical processing unit and quantum memory".

But then I've only just seen a simulation of 50 qubits using a supercomputer. https://phys.org/news/2025-11-full-simu ... antum.html

There isn't really such a thing as quantum memory, but I suppose you could have slow qubits with long coherence times as storage while you do your computation with fast qubits that have short coherence times. (You can't just "overwrite" the memory with the qubit state you want to store, you'd have to reset the memory qubit to a known state and then swap your qubit state with the memory state.)

[shpalman]

... This just got referenced: https://sam-jaques.appspot.com/quantum_landscape_2023

You can actually look at the 2021 and 2022 updates to see the progress. And then remember that the graph has logarithmic axes.

And we can now look at https://sam-jaques.appspot.com/quantum_landscape_2025



There hasn't been progress in where we are, but there's been progress which has brought the requirements to break RSA down.

The big news: the curves on the right (the resources needed to break RSA) have moved by a factor of 20, meaning we only need one million physical qubits to break RSA-2048, thanks to a new paper of Craig Gidney.

To put this in perspective, this brings a quantum attack as much closer to reality as if someone built a 2,100 qubit device this year, and probably more (since we can now avoid any unforeseen scaling difficulties between 1 million qubits and 20 million qubits).

...

If we're imagining an exponential growth in quantum resources (which we haven't seen yet, but we need to if we expect to ever reach 1 million physical qubits), this shaves off 4.3 doublings. Today's physical devices are "only" a factor of 9,500 away (13.2 doublings) from this new target.