LLM generated UI for MVPs and explorations seems acceptable, but I don't read every Show post (maybe I should!). But when tinkering becomes a product it should have its UI revised when starting to take it seriously -- human touch for Human Interfaces pays off (even if AI augmented in the effort).
The other issue of HN being inundated with AI bots is related, but a kind of different problem.
Intel used to say, "For every GPU sold there's an Intel CPU sold." At least in the PC gaming days, when that was Nvidia's core biz and Intel had no real response (frankly, that apathetic retort could explain a lot about Intel's struggles). Now, Nvidia is taking a similar approach -- for every QPU sold there are n GPUs sold. And, they will be right for the next 5 years or so because hybrid workflows, where classical are integral to the workload and quantum is specialized, are the common case. But, if QPU roadmaps are true, GPU/CPU supercomputers will be losing ground to QPU supercomputers. At which point, Nvidia probably acquires a QPU company. And QPU companies merge with other GPU/CPU companies. Disruption will force a change in the supercomputing business. Nvidia may or may not be a winner. We shall see.
This is a pretty impressive development in quantum. Neutral Atoms will likely be a cornerstone modality in qc for the foreseeable future. The error correction regime is driving so much advancement. Prof. Preskill's role should also be taken seriously, as he has historically leaned very heavy into academia and not into commercialization. Even JP has gotta get his bag
Commercialization can bring in speculators and hype. And, I'd argue that speculation is a necessary for accelerating market development. Commercialization brings with it unique forcing functions that don't exist in academic settings, and this historically leads to acceleration of functional products. The first step is building a quantum computer to learn how to build a quantum computer. That step is done, while research continues in many areas, the commercialization challenges are largely engineering in nature.
I've only seen 34 qubit simulators (eg AWS SV1). My understanding is that 34 qubit uses 512GB of RAM, and each additional qubit doubles the RAM requirement. So, 50 qubit simulated would require 16.8M GB of RAM.
100 logical qubits seems to be the minimal threshold for interesting/useful quantum computing, albeit with very limited use cases. Classical still beats most. Quantinuum will hit that number in 2027. And, IonQ (often cited as being a hype-machine) expected to have 800 logical qubits in 2027.
The industry is moving out of the NISQ Era (noisy-intermediate-scale-quantum) and into the Fault-Tolerant QC (FTQC) era. NISQ is experimental. FTQC is commercial (ie reliable, repeatable).
Nice, an informative meaningful comment. From your userid; would i be correct in deducing that you are affiliated with QuBOBS Project? - https://qubobs.irif.fr/portfolio/
You are certainly right that commercialization (and speculation does play an important role here) serves as a forcing function to accelerate development of products. But this needs to be done somewhat in-sync-with/a-little-ahead-of the actual science and engineering. When the subject is inherently difficult to understand (as is the case with QC) it can very easily get out of hand and become just snake-oil/bullshit and exploited by hustlers/grifters/charlatans.
Do you have any links to more information on the points that you make above that you can share? Specifically on hybrid quantum-classical systems and silicon-based shuttling-qubits which can use current foundry technology? To me, this seems to be the future since both the scaling and availability are taken care of.
Thank you :). I'm not affiliated with Qubob project, it's just a name I picked.
QC has had its share of hype cycles. Businesses need to have a vision (which is where hype seeps in), and must be honest about what's possible today.
I am building quop.ai, a service which makes quantum computing accessible to technical and non-technical folks who don't have degrees in quantum physics.
Thanks for the pointers. I had seen them before except for the qutech one. I firmly believe the silicon approach to qubits is the way to go; but we will see how the market (and technology viability) settles everything. These are exciting times for hard Physics.
Also had a look at your quop.ai; seems pretty interesting though i need to explore it a bit more.
You might want to think about posting quop.ai to HN and get some feedback ;-)
Exciting times indeed! Will be fun to see how this all plays out.
Thanks for taking a look at Quop. If you explore it more, I'd be curious what you think as someone who follows the hardware side closely. I'll post something soon.
Also, there's a CalTech affiliated startup called Oratomic Inc that is building a Rydberg system. I think it's the same team that demonstrated 6100 qubits. And, I hear Preskill is an advisor.
The other issue of HN being inundated with AI bots is related, but a kind of different problem.