"This scheme probably will not work"

[leblon]

Интересная критическая статья о "квантовых компьютерах" написанная физиком. Я, правда, пафоса статьи не разделяю. Все эти пугающие экспоненты там именно для того, чтобы напугать читателя, реального смысла в них нет.  

Comments

[juan_gandhi]

Сейчас поеду в город буду слушать два доклада про "квантовые вычисления". В глаза хотя бы этим людям поглядеть.

О, и оказывается Ю.И.Манин идейку подбросил! А потом уже Фейнман.

(Interesting how the only language people can use describing all this are "sets" and "continuous".)

[alexanderr]

будут они в итоге работать или нет, на самом деле не важно. важно, что они обещали code breaking. т.е. это гарантия того, что государство их будет финансировать неопределенно долго. ученые нашли способ гарантировать интерес и поддержку к своей безнадежной теме исследований. respect. т.е. смысл этой статьи следующий: "пожалуйста, дайте еще больше денег, т.к. тема очень трудная". а обратный эффект (закрыть тему) невозможен. т.к. всегда есть опасения, что враги напрягутся и тайно сделают и будут взламывать коды.

при этом уже давно есть вещи, которые работают на коммерческом уровне. но это конечно не computing, а quantum encryption. вполне себе действует, и даже летает по самому обычному fiber'у. или с китайского спутника.

[chaource]

Bill Unruh wrote a paper a while ago about the problems with quantum computing.
https://arxiv.org/abs/hep-th/9406058
These questions are hard, and probably most researchers aren't spending a lot of time answering them.

My take on the issue is that 1) D-Wave devices are not what was usually called quantum computing - they are not general-purpose quantum computers and, for instance, can't execute the Shor algorithm or other such algorithms. D-Wave devices are analog computers for special purpose computations. 2) Quantum computers have an unsolvable tension between trying to keep all qubits as isolated as possible to prevent errors and decoherence, and at the same time keep all cubits connected to a large number of classical control devices, so that the quantum computer can reliably execute all the different quantum operations, such as a phase rotation across a given subset of cubits and so on.

I believe that this tension is unsolvable, in particular, because designing a 10-qubit quantum computer is a completely new engineering challenge and not reducible to connecting two 5-qubit computers together (as is the case with classical computers). I heard a talk some years ago where it was explained that the design for 5-qubit computer does not work with 6 qubits, and different atoms need to be used as quantum devices for 6 qubits to work, and yet other atoms for 7 qubits. The periodic table has only a few atoms with specific necessary properties, and very quickly the designer of a quantum computer simply runs out of suitable atoms, so a 5-qubit design cannot be extended to 10 qubits.

If increasing the number of qubits by 10 requires a whole new design, it's safe to say that we will never get 1000-qubit computers.

[leblon]

I am by no means an expert on this. But a lot of time has passed since Unruh's paper, and much effort was spent on understanding how to fight decoherence. One big idea was topological quantum computation, which uses the locality of interactions to design qubits which are much less sensitive to decoherence. I also recently saw a talk which discuss possible designs based on Majorana qubits, and they looked pretty uniform in the number of qubits. Admittedly, topological quantum computation is less advanced than other attempts, but it is quite promising. Here is the talk: http://pmaweb.caltech.edu/PhysColl/abstracts/18-19/vonOppen18.html . It is mostly theoretical, so one might ask whether these Majorana fermions can be really realized. There is a little bit about it in the talk, and a more complete discussion of the recent progress with Majoranas is here: http://pmaweb.caltech.edu/PhysColl/abstracts/18-19/Yazdani18.html . So while I am somewhat skeptical about the final success of all these schemes, I would not dismiss this very interesting work as a mere attempt to get more money from the government.

[chaource]

Of course, we can't dismiss the research just because it was funded by the government. There are two conflicting forces here: first, if the government gives money for research and does not demand immediate practical results, there will be some more good research with long-term focus. Second, there will be more research that has no practical results and merely a claimed (but incorrect) theoretical possibility of something to be developed in the future. The Majorana fermions might be impossible to use for constructing a quantum computer, due to some unforeseen difficulty, but the scientist is not motivated to examine that difficulty because there is no requirement of delivering an actually working device, and because of the tendency to prefer publishing positive results to negative results. The consequence - unintended, of course - is that a group of scientists will start publishing purely theoretical papers on various theoretical aspects of using Majorana fermions for quantum computers, while none of these scientists are actively looking at any practical aspects of building quantum computers. So these scientists are likely to collectively miss some crucial difficulty that makes Majorana fermions unusable. They will all unconsciously avoid studying any of the topics related to this difficulty. This research domain could be "flourishing", in the sense of growing number of papers and citations, and yet will lead to no practical outcomes. Maybe some experimentalist would build a 20-qubit computer with Majorana fermions, discover some unsolvable problems with the 25-qubit design, and move on. But this kind of experimental setback was never a deterrent for theorists to stop publishing on their chosen topic. Eventually the theorists will run out of theoretical questions that can be answered within 6 months and result in a publication, and will move on to another topic. The practical problems with Majorana fermions will remain undiscovered and unsolved, but occasional graduate students will continue to write theses on the topic. I'm not saying that this is indeed the case for Majorana fermion quantum computers, since I have no idea about that, but this is a likely outcome of government grant-funded theoretical research today.

Majorana fermions or something else might be indeed new mechanisms that are less sensitive to decoherence. However, the central difficulty of quantum computing still remains - you need to have a large quantum system that is at once well-insulated from the classical world to avoid decoherence, and yet at the same time all qubits must be strongly coupled to some parts of the exterior classical world so that the quantum state of the qubits can be controlled by the quantum computer's program code, executing an arbitrary sequence of quantum operations on a number of chosen subsets of qubits.