The most common example I’ve heard of for application of QC is national security applications for breaking encryption codes. That said I could see it having a potential impact on high frequency trading. But for a swing trader who’s not trying to necessarily compete with speed I don’t see how it would have an impact.
HFT wise, normal high end computers more than fast enough for the job, the issue is the data ping times, not the CPU.
Depends on what you define as normal high end. Speed of the CPU does matter in hft. 5Ghz+ single core CPU performance is what most top hft firms employ these days.
Not a lot of data to crunch, mainly just looking at Level 2 data and front running orders for tiny profits. I'd bet wayyyy under 1ms per trade. Also, who's to say Quantum could actually perform that kinda task.
I never stated what you put in my mouth. I said 5Ghz CPUs are used among the top hft firms right now. And yes, they play in the micro second space
Lots of misinformed opinions on quantum computing. I am a computer scientist by education and trade so I'll try to enlighten some of the more bad ones. My specializations are computational mathematics so I only have a graduate-class level view of quantum tech. This is generally correct. The size of the machines and technology currently required is very limited but promising. Quantum technology is already used, and is one of the only true random number generators on the market. This isn't really "quantum computing" but more using the strange properties of matter at that scale to create a truly random event to sample. One of the more meaningful, direct, applications of quantum computing is Shor's Algorithm which will render all modern cryptography worthless. Accurately predicting one minute ahead isn't kneecapped by computation it's kneecapped by the fundamental properties of probability. However, as for AI, the general consensus in the applied fields is that we are at a computational cliff currently where deep neural networks are simultaneously impossible to understand and very difficult to compute. Complicated matrix operations (the kind you find in advanced probability theory) on classic architectures have generally intractable big O values. A good example is the laplace expansion of a matrix which has a run time of O(n!), basically impossible to compute quickly. Whereas quantum computing will not reduce the mathematical upper bound of complicated matrix operations, adding horsepower to each calculation will still provide some tangible improvement. This leads directly to solving problems in AI faster, such as stochastic gradient descent. I'm certainly onboard with "complicated math problems are not true AI", however I will not be surprised if we have something close to in the next few decades. We have already simulated rat brains and currently the fundamental limitation of this research is the size of the data and scale of the computational power required for more complicated creatures. We are actually limited almost entirely by our classical computing architecture and quantum computing holds a lot of promise in solving this. This has not been what I have heard. ASICs are deployed regularly by many firms (Jane Street is one of note) because modern general hardware is too slow.
Some do deploy fpgas but the majority don't. I like your post above, quite informed, factual, and well written. Thanks for sharing your thoughts.
So according to Gaussian quantum computers can possibly operate in a fundamentally different way to current computers , it's not just speed?
They only have like 64 Qbits or something equally as stupid for huge cost, what can you do with so few? And the normal cpu with normal code has to control that. That's why i say very specific small functions with very limited real world use is faster only.