Setting aside algorithmic conventions, the only reason that there are still domain experts who will look you straight in the face and tell you that quantum computers [1] will never outperform classical computers at all is because they are often of probabilistic utility where classical are often deterministic. However classical computers, when you account for their exposure to edge case scenarios, are also probabilistic. Just of a different distribution…
The question of whether or not quantum computing dominates classical resources (“quantum advantage”) isn’t just a question of probabilistic distribution characteristics. It’s also a simple matter of having different regimes of application scales where the resources become appropriate. Ideally quantum computing systems should be reserved for when the available hardware becomes capable of addressing the largest scaled settings, where their strengths may be particularly impactful, not just in latency but surrounding the significantly important implications of Carbon Intensity of our economy’s scaled compute (not to mention surrounding infrastructure costs). My own expectation is that the best path to market for future quantum computing vendors will be to focus on what new business models may arise when performant optimization becomes accessible to scaled systems in new settings.
This isn’t a scientific observation but kind of my own expectation that the probabilistic nature of solution sampling may be more tolerable to human participants in those scaled systems as well. A human trying to operate around probabilistic algorithms who isn’t trained to understand what it means that algorithms are operating probabilistically may misinterpret sampling from different modes as some form of indication of failure. The reality is that for any domain of uncertainty in which multiple candidate trajectories are accessible those different directions of information are actually helpful, no matter how confusing it may be to those accustomed to single minded classical alternatives.
The reason that commercializing pure play quantum computing firms was a nearly intractable Problem from the very beginning is that widespread convention of modern business practice is that you have to prove yourself in an early use case before getting access to scaled settings. Kind of like how GPUs became viable first as video game engines. That simply doesn’t work in quantum computing where the systems have a fundamental disadvantage in low scale settings (where deterministic algorithms can be expected to dominate).
In quantum computing, the route to viability isn’t to establish a scalable business model that you can then fund and scale out in the traditional venture capital (VC) sense. The only realistic way to demonstrate the value is to apply to a business domain that already has established access to a scaled Setting. The fact that pure play vendors can only realistically demonstrate value in already established scaled settings means they have almost no leverage to negotiate with established big tech (beyond those sources of leverage that are unethically established). After all there are no shortage of substitutable qc vendors that can be directly applied if an established resource gets out of hand. Which is kind of how it should be. (Although short term advantages originating from hardware or software may be realistically expected between different conventions from time to time, those advantages should reasonably be channeled to higher clearances where they belong.)
There will always remain a need to evaluate how to prioritize our allocation of non-infinite computational resources. The purposes of proxy competitions isn‘t to determine who has the biggest balls. The alternative to healthy proxy competitions is unhealthy direct competitions. Traditionally this is what markets have been for. The hard part then becomes how we can structure markets in a manner that effectively adapts to the new reality of where new sources of abundance and changing bottlenecks have changed the behavior of markets and the economy. The more critical the need to source serious information of this nature, the more proxy competitions may start to appear less of a silly nature…
The challenge has always been and remains that benchmarking competitions are only informative to the extent that the patterns aren’t based on single sampled outcomes but based on endurance over prolonged time periods and further with an expectation for potentially changing results with evolving distributions which may be linearly or nonlinearly transformative. The only realistic way to incorporate this is with HUMAN intelligence. That is why peer reviewed research publications and experimental findings and etc still matter. Because until we literally have established a theory of Everything that we can effectively extract towards future states, any realistic expectation of sustainability in a competitive ecosystem demands that we will need to channel the finite intelligence at our disposal to do way more than just tell AI agents to transform lead into gold…
The ethical challenge then becomes that proxy competitions are less desirable when participants don’t know the basis of competition or why there are a part of it. And if a party is drafted into the domain without such a courtesy introduction there should be an easy out made available, whether by a letter of resignation or perhaps just a reasonable transaction of sale. Ideally we want to carry forward as much of the legal scaffolding that originated from the pre AI era that has already been proven to effectively operate in settings with exposure to diverse legal regimes and complicated trade imbalances. If we try to just rebuild everything from scratch that is not likely to resolve quickly…