This is basically a tautology and true for any piece of programmable technology. But, of course, Intel wants to drum up some PR around (quantum-related) software they recently announced.
For the time being, without a doubt, hardware is the gating factor to the success of quantum computers. Theorists and engineers have built a ton of algorithms and software already, waiting to have hardware that performs even just barely sufficiently to handle them.
Yeah this article is a nothingburger of Intel still aspiring to software margins. I wish they could just own it and be happy to make cool reference designs and go back to winning at hardware.
Machine learning is solving more and more of the optimization problems that quantum computing was ,,supposed to'' solve. AlphaFold is a great example as it is getting better every year at simulating biochemistry.
The only important problem set left is beaking cryptosystems which gives governments even greater power over us.
Not sure why this is getting downvoted, I think it's a good insight. There's often more than one way to solve a problem, and ML has shown that one can solve certain problems by extracting commonalities from existing solutions, without understanding the true mechanism below. AlphaFold showed that we can "solve" protein folding without solving the underlying quantum mechanics, which is something that can probably be done in other areas as well.
That said there might be some problems that might only be solvable using a full quantum simulation, though personally I think that class of problems will be rather small.
If I am not wrong, some of the probelms AlphaFold can solve are quite limited. Even if the results are promising, protein folding is an insanely complicated problem and having tools like alphaFold be good at some aspects of it can be useful but cannot crack the bigger nut by design as far as I am aware.
It currently performs ok on protein prediction but cannot do new proteins, which would theoretically be possible with quantum computers. And its not just a matter of giving alphafold more resources they are entirely different problems
I don't totally understand your point. AlphaFold is great at protein folding, and the main limitation is predicting protein structures with multiple states when protein interactions happen. That's the next task the team is working on (it's the natural next step for biology simulation), and if the team is successful, it will be incredibly valuable (it can easily be the next multibillion dollar business for Alphabet, and a huge improvement in how drug research is done.
I don't see quantum computers anywhere near as close to solving a real problem with drug research as AlphaFold (although I don't argue with the theoritical possibility in the future, but not in the next 5-10 years, where AlphaFold can shine).
I apologise as my field is not related to proteins, but I do recall an interesting read about the subject from a few months ago. Here is the link if you wanna follow it yourself.
There are some interesting criticisms from the language, to the results, to the limitations of AlphaFold that I think seem to represent the prevailing opinion from biologists about the current scope of what ML can achieve on their field.
There is also push back to the complaints and some very optimistic people about Deep learning models on the field so perhaps you are right.
On the other hand the governments benefiting most from strong public key cryptography enabling cryptocurrency are dictatorial ones like North Korea using it to launder past sanctions. For private personal communications you can use symmetric or one time pad since you can exchange keys/pads in person and quantum hasn't been shown to put that at risk (and even has enhanced flavors of them that are even more secure).
North Korea launders most of its money through US banks (billions) compared to tens of millions going through cryptocurrencies.
I worked very hard for many years and payed taxes legally and bought Bitcoin 100% legally through regulated exchanges so it's quite boring to me that money laundering comes up all the time. Most of the money that governments steal are made by tax abiding hard working citizens, not dictatorships that destroyed all their wealth.
Software is what makes hardware useful. I see this with a lot of IOT companies I deal with. Great hardware but the software is often a mess and an afterthought. So, you end up with users of that hardware either having to write a lot of software themselves or being severely limited and restricted by the software that comes with the hardware.
Quantum computing is not going to be very useful until normal programmers can work with it. If you need a Ph. D. in theoretical physics to even understand what is happening, there are not going to be a lot of people that qualify. I don't see this happening very soon.
Right now quantum computing has three problems:
- few people even understand what it does and fewer are capable of explaining that in a sane way. Articles on this topics are using a lot of quantum computing cliches and bad analogies (cats in boxes, etc.) to make it seem like it's some kind of dark magic that will solve any problem trivially. It does not help that professional snake oil sellers like IBM are all over it either.
- Current hardware is not very practical, useful, or cheap. Not enough qubits to do much useful things, lots of issues related to error detection and mitigation, not scalable.
- The only quantum computer software that exists is essentially proofs of concept and research prototypes. Basically it shows that if we had a working quantum computer, we might be able to make it do some useful things. But nobody has gotten there just yet.
Before software, hardware had more limited usage scope. A music box can play any song, provided the right "file", the same way an analog computer can simulate anything within the capabilities of its hwardare, with the right set of wires and dials.
I agree and that is why we have build Norse https://github.com/norse/norse an open source deep learning library for neuromorphic computing based on PyTorch. It doesn’t have any commercial backing yet, but we had quite a bit of interest in from other research groups and it predates Lava by quite a bit .
For the time being, without a doubt, hardware is the gating factor to the success of quantum computers. Theorists and engineers have built a ton of algorithms and software already, waiting to have hardware that performs even just barely sufficiently to handle them.