First I heard of these exascale computers was when a Chinese group published[1] a pre-print in which they classically replicated the Google quantum supremacy result in about 300s. The Google team estimated that a supercomputer (Summit) would take 10e3 years to do the task, which apparently really got to the classical HPC community.
If you want to import anything at all you need foreign currency. One of the ways you can do this is by exporting goods, of course, but you could also sell some of your local currency in exchange for dollars or whatever. Bitcoin volatility makes this harder. Also I’m guessing that volatility makes it hard to get foreign currency loans collateralized by BTC
I’m of a similar opinion, but with the note that QCTRL actually makes products that people want. The current state of the art in QC requires that pretty much everyone needs some kind of pulse shaping/noise reduction, which is (some of?) what QCTRL provides. Plus the quantum sensing space benefits greatly from such techniques as well. So even if quantum computing falls into a “quantum winter” it’s likely that quantum sensing will pull through and so QCTRL might survive.
Those other software companies are also competing with the IBMs, Googles, and Rigettis that are also building similar software for the the hardware they’re making.
I use blogdown, which handles a lot of dumb bullshit that comes with static site generation. Also supports latex, markdown and you can use RStudio to create and preview posts.
The apparatus you’re thinking of is called a dilution refrigerator.
The “tiny pipes” are actually coaxial signal cables carrying RF and DC control signals. These terminate at a “package” which contains the chip on which the superconducting qubits are located.
The disks (we call them plates) are different temperature stages for the refrigerator, with lower stages being colder than upper stages. The last stage has operating temperature ~10-20mK. Then 100 mK, 1K, 3K, 50K and lastly room temperature. Most of the other things on the refrigerator are involved either in signal filtering, thermal anchoring, or thermal isolation. When operational, all of the stages are enclosed in multiple cylindrical thermal shields. The outermost layer is the “vacuum can” which is airtight and allows for the whole thing to operate under internal vacuum.
Edit: This is just one type of quantum computing device. Others, such as trapped ion or neutral atom rigs would look radically different. Larger vacuum enclosures, different ways of performing control/read operations (lasers!), etc.
This is true for now. The IARPA SuperCables program exists to tackle this specific problem. Probably the solution will require converting microwave pulses into optical frequencies to get the signals out of the cold spaces with fiber-optic infrastructure.
If you read the broad agency announcement, they even have explicit energy dissipation targets (per bit). I think it’s something like attoJoules or femtoJoules?
They’re clearly pretty serious about it (why build a gigantic dilution refrigerator, otherwise?) but they’re definitely going to need to wring a factor of two out of the precision of their device fabrication process. Specifically, the laser annealing process for their Josephson junctions currently leaves their qubits with about a 14 MHz frequency spread. The IBM approach to this requires that the qubits are fabricated very precisely in three well-separated frequency bands. I’m pretty sure any frequency “collisions” between qubits in different frequency bands make them useless. Right now they’d have about an 8% chance of fabricating a 127 qubit chip with no collisions. To get the same chance to yield a 1000 qubit chip, they need qubit frequency spreads between 9-10%. The main constraint on this is currently the difficulty of predicting qubit frequencies (measured at milliKelvin temperatures) from resistance measurements conducted at room temperature.
I’m hoping they’ll publish when they figure it out, because I’m really curious about how they’re gonna crack this one.
Have you considered industry research positions? Depending what field, there could be high demand for your skill set, even if the research is not exactly what you did in your academic life.
I went to industry immediately after grad school and it has been pretty dope. It might feel differently to you, since you spent so much time in the academy.
1. I write mostly about happenings in quantum information and quantum computing. Ranges between dives into interesting papers and commentary on recent funding/start-up developments.
2. Whenever I feel inspired. Sometimes that means several posts within a week or two, sometimes it means months without posting. I realized I can’t force myself to do it unless I’ve got something on my mind. I just can’t write about something that I’m not obsessed with.
3. I host it on netlify as a static site. Seemed easiest at the time.
There are annealers that have thousands of qubits (DWave), but they are low coherence and very different from the gate model qubits in use by Rigetti, Google, IBM, which are themselves lower coherence than the ions being used by IonQ and Honeywell.
It’s all very much in the research phase so far. AFAIK there’s no reason to pay for time on any of the available systems (all <100 qubits unless you’re using a DWave annealer) unless you’re doing basic quantum computing research or benchmarking. Folks are working hard to make the small, noisy systems we have now do something useful, but the real moneymaker will be error-corrected and fault tolerant qubits. Those may be available perhaps between 5 and 50 years from now, depending on who you ask.
The Serious Work right now is to build scalable quantum architectures and continue to pursue error correction and fault tolerance. There’s also a shitload of work on optimizing control, devising classical architectures that will mate to the QPU, and so on.
Unfortunately, this announcement is mostly just moving around some of the 1.2 billion earmarked for the National Quantum Initiative back in 2018 or so. Unclear to me if any new money is being added to the pot.
They could using a small EMP weapon. These happen to have short range and are explosively pumped, so at that point it’s easier to just disrupt the magnetometer by initiating rapid explosive disassembly.
My understanding is that it doesn’t have to absorb energy to perform calculations. The reversibility dependent on keeping track of all of the correlations in the system.
[1] https://arxiv.org/abs/2110.14502