What is the danger of using 'quantum volume' as an indicator of quantum computer performance?
Shtetl-Optimized »Blog Archive» Turn down the quantum volume
On March 2, 2020, Honeywell , an American multinational company engaged in business in a wide range of fields from thermostats to aviation and military industries, announced that it will develop a quantum computer with `` quantum volume 64 '' and release it by mid-2020. did. IBM states that it achieved quantum volume 32 with its quantum computer in January 2020, and Honeywell claims 'at least twice as powerful as traditional quantum computers.'
Honeywell develops and announces QV64 quantum computer that is twice as powerful as before. Released by mid-2020-Engadget Japan
Honeywell claims that 'in the next five years, we will increase the order of the quantum volume by one every year,' and has been spotlighted as potentially bringing significant innovations in the field of quantum computers. Aaronson commented that his friend contributed to Honeywell's quantum computer development and welcomed Honeywell's entry into the field.
'I would like to comment here on the details of Honeywell's announcement: focusing on' quantum volume 'as a central indicator of quantum computer progress, and Aaronson said,' On the basis of what we call a device the 'most powerful quantum computer'. '
After Honeywell announced his quantum computer, Aaronson asked journalists to ask, 'Why is quantum volume important as a quantum computer index?' Aaronson responded, 'I do not know if quantum volume is really important,' and questioned the use of quantum volume as an indicator of performance.
In the first place, quantum volume is a criterion proposed by IBM.It combines various indicators such as the number of qubits and error rate, qubit interconnectivity, qubit fidelity and conductivity, and improves the performance of a quantum computer by two. It is expressed as a power of. It seems unclear why IBM represents quantum volumes as powers of two, but Mr. Aaronson points out that the aim of quantum volumes is to `` show the performance of quantum computers that can not be easily quantified with easy-to-understand indicators ''. .
Regarding the idea of communicating the performance of quantum computers to consumers in a way that is easy to understand with quantum volumes, Aaronson admitted to some extent that `` quantum volumes seem to be a clear step in the right direction ''. . On the other hand, Mr. Aaronson brought out
An example of where Goodheart's law can be applied is Aronson's ' h-index, ' which shows the relative contribution of researchers based on the number of papers and citations. The h index is used when considering the hiring and promotion of researchers, but the problem is that `` the h index can only be used for comparison between researchers in the same research area because the citation practice differs depending on the research field '' A point has been pointed out, Aronson points out, that the life of the researchers is 'regularly polluted' by the h-index. George E. Hirsch, the creator of the h-index, also expresses regret about creating the h-index.
Aaronson fears that quantum volume could also be an example of where Goodheart's law applies. As quantum volumes are generalized as indicators of the performance of quantum computers, many companies and research teams have been focusing on creating `` quantum computers that exactly clear only the evaluation items of quantum volumes '', The original purpose of 'realizing performance exceeding conventional computers by incorporating quantum theory' may be damaged.
If certain metrics are set as goals, 'the only' highest score 'can be achieved by exploiting the flaws in the scoring system,' Aaronson said. For these reasons, Aaronson is concerned about using quantum volumes as a measure of quantum computer performance.
Therefore, Mr. Aaronson proposed as an alternative, 'publish the specifications of all systems as clearly as possible.' In addition to items such as qubit count, interconnectivity, and fidelity, the most important information is 'the main shortcomings of the system, the worst parts, the areas that need the most improvement, why you can't use a quantum computer at this time.' Aaronson, arguing that at this time it should not be combined into one indicator.
in Science, Posted by log1h_ik